JP4733835B2 - High-speed bit swapping in multi-carrier transmission systems - Google Patents

Description

【００６４】
ここで、Ｈ_nおよびＮ_nはチャンネルゲインおよびノイズパワーである。（事前の初期化、再トレーニング、または高速スワップをもとに）事前に識別されたチャンネルゲインのセットは、ＤＭＴシステムが事前に識別した各チャンネルプロフィルために格納することができる。これらの格納されたプロフィルは、ｌ＝１，...,の場合にＰ_l,nで表される。ＤＭＴシステムクロックは、プロフィルが最後に生じたときから相対的なタイミングオフセットεだけずれるので、古いプロフィルの設定から線形に位相偏移を生じる。
【００６５】
【数２】

【００６６】
首尾良いプロフィル検出によって、事前のプロフィルのなかに、同期シンボルの既知のＸ_nに対して最後に測定されたＨ_nに最も良くマッチするプロフィルが存在するか否かが識別される。
【００６７】
先ず、プロフィルは既知でないので、プロフィルを生成しなければならない。Ｈ_nの単純な推定値は、数３である。
【００６８】
【数３】

【００６９】
ここで、Ｍは使用された同期シンボルの数である。Ｍが大きいほど正確なチャンネル推定値を得ることができるが、訂正高速スワップの実現は遅れる。推定値の差異はＭに比例して減少するのが一般的であるので、測定されたＳＮＲは、９９％の確定度で、数４だけ高すぎるまたは低すぎる（高すぎると問題になる）。
【００７０】
【数４】

【００７１】
４つの同期シンボルを経ると、算出されたＳＮＲエラーは９９％の確定度で２ｄＢ未満になる。あらゆるレシーバゲイン／位相回転（一般に「ＦＥＱ」と称される）を、数５の比で直ちに調整する必要がある。そして、もし必要ならば、新しいチャンネルをプロフィルの１つとして格納することができる。
【００７２】
【数５】

【００７３】
しかしながら、あるプロフィルが事前に識別されていた場合は、このプロフィルの再発生を最大尤度のプロフィル検出に従って高信頼度で検出するために、同期信号を単独で使用することができ、こうして、（可能なベクトルプロフィルＰ_lに対する）和を最小化するプロフィルが選択される。
【００７４】
【数６】

【００７５】
偽のプロフィルが検出される可能性は、ノイズおよびプロフィル間の近さに依存し、特に、数７による制約を受ける。
【００７６】
【数７】

【００７７】
ここで重要なのは、互いに混同された２つのプロフィルが近すぎて、対応するビット／ゲインテーブルが機能的である場合は、偽の検出はリンクの機能異常に対応しないことである。レシーバエキスパートシステムによってプロフィルを厳密に維持することによって、対応する訂正ビット分布がＤＭＴのリンク故障に通じる偽の検出を生じる可能性が非常に低いように、プロフィルを選択する必要がある。
【００７８】
さらに、パワーの合計測定値によって顕著なチャンネルゲイン（または出力パワー）の変更が検出されなかった場合は、顕著なノイズの変更が想定される。ノイズの差異を正確に推定するためには、トレーニング信号を使用した場合であっても、ＤＳＬモデムにおいて数秒の時間を要する。しかしながら、通常のように、ノイズ導入前のマージンが高かった場合は、サブチャンネルを慎重にロードすることが可能である。ガウスのノイズサンプルは、０．９９７の可能性で±３σ（３標準偏差）分布点の範囲内に入るのが通常である。ノイズが１０ｄＢまたはそれ以上増加すると、６ｄＢのマージンで１ｅ−７でもともと動作しているＤＳＬモデムにおいて、許容不可能なほど高い誤差率が引き起こされる。レシーバ内のＦＦＴ出力におけるスライサーエラーの有無に拘らず、大きいノイズに対するエラーサンプルの分布には、より大きな差異が存在する。レシーバは、一般に、その分布を推定することができる（ガウスノイズに関しては、この分布によって差異が推定される）。新しいノイズの厳密な差異を非常に正確に推定するためには相当数のサンプルが必要である（１／１０ｄＢの精度を得るためには３２００程度のサンプルが必要であり、これは、非効率的なエスティメータを使用した場合でも１秒以内のトレーニングですむ）一方で、大きな変更に対応してより高速なグロスインジケータを使用することも可能である。顕著なノイズの増加が検出されたこれらのトーンは、ＥＳを介してビットテーブル内でゼロ化される。これらのトーンがＥＳ要求に応じて非動作されるのに伴って、後続のノイズの推定を継続することができる。ローディングの結果がより正確である場合は、後続のＥＳコマンドによって、より詳細化されたビット分布を伝送することができる。さらに、ＩＳＤＮクロストーク、ＨＤＳＬクロストーク、または他タイプの既知のクロストークに近いノイズプロフィルを最初に概算し、次いで高性能のレシーバで詳細化することによって、ＥＳコマンドを介した非常に高速な調整が可能になる。
【００７９】
さらに詳述すると、ノイズの変更が大きい場合は、何の信号も伝送しない、すなわちトーンから０を出力することが、ノイズを決定する効果的な方法である。以下では、トーンの出力ゼロ化によってノイズを推定する方法を幾つか説明する。
【００８０】
一般に、チャンネルの変更がなくて（仮定）同期シンボルが既知である場合には、各トーンの瞬間的なノイズサンプルを決定することができる。これは、数８によって最適に表される。
【００８１】
【数８】

【００８２】
ノイズパワーは、数９によって推定される。
【００８３】
【数９】

【００８４】
次いで、レシーバは、これらのトーンをゼロ化するよう、ＥＳコマンドを介してトランスミッタに通知する。ここで、数１０は、もとのノイズパワーに比例した特定の閾値より大きい。
【００８５】
【数１０】

【００８６】
一般に、ノイズの推定を高速化するためには、レシーバが、小さい値をＭとして選択しなければならない。すると、ノイズパワーの推定が不正確になり、このためにトーンの出力ゼロ化のプロフィルが不正確になる恐れがある。トーンの出力ゼロ化のプロフィルの精度は、ノイズ差異のパワースペクトル密度が比較的滑らかであると想定することによって、改善することが可能である。このような滑らかさを想定することを利用したクラスタ化の方法は、数多く存在する。
【００８７】
１つの実施形態において、連続したトーンセットをクラスタ内で規定する状況は、一般に、ゼロ化されたまたは未使用のトーンであるエンドトーンであり、クラスタに先行するおよび後続するトーンは使用後のトーンである。この実施形態では、モデムによる使用を意図したトーン（すなわち非ＰＯＴＳトーン）のみを対象とする。一般に、ゼロ化トーンは、ノイズが充分大量に増加したために０ビットを伝送するトーンとして規定され、使用後のトーンは、新しいローディング後も非ゼロビットを伝送するトーンとして規定され、未使用のトーンは、チャンネルの低ゲインまたはノイズの高差異のいずれかが原因で、ローディング前にゼロビットを伝送していたトーンとして規定される。
【００８８】
以下に上げる実施例において、クラスタ化の方法を示す。ここでは、記号０、Ｘ、Ｕが、ゼロ化トーン、使用後のトーン、未使用のトーンをそれぞれ表すものとする。２０トーンに対し、次のようなトーンの出力ゼロ化プロフィルが与えられると、
【００８９】
【表１】

【００９０】
可能なクラスタは、（２，２０）、（２，１８）、（２，９）、（２，６）、（４，２０）、（４，１８）、（４，９）、（４，６）、（８，２０）、（８，１８）、（８，９）、（１２，２０）、（１２，１８）である。
【００９１】
クラスタの識別に続く次の工程は、どのクラスタの非ゼロ化トーンをゼロ化するかを決定する工程である。これは、以下の経験則に従って行うことができる。先ず、少なくとも一定割合のトーンがゼロ化された最大サイズのクラスタを探し、次に、同じ割合のトーンがゼロ化された次に大きいサイズのクラスタを探し、これを、全部のクラスタが探されるまで繰り返す。このクラスタのサイズが使用後のトーン数と同じである場合は、リテインする。それ以外の場合は、クラスタ内の非ゼロ化トーンをゼロ化し、もし可能ならば使用後の他のトーンにビットを移動させ、すでにゼロ化されたクラスタは無視して第１の工程に戻る。上述した実施例において、上述した割合が７５％である場合は、トーン３、７、１９がゼロ化される。
【００９２】
ＥＳコマンドを送信する必要なしにトーンをゼロ化する代替の方法は、トランスミッタおよびレシーバにとってともに既知である所定のシーケンスを使用して、トーンのサブセットをゼロ化するものである。この場合は、このトーンゼロ化の方法の開始を合図するコマンドが必要である。測定される信号がノイズのみであるゼロ化されたトーンでは、ノイズパワーの移動平均推定値を得ることができる。ゼロ化されたトーンに関して、信号の平均が４０シンボルを超える場合、信号は、９９％の信頼性で真のノイズ差異の２．０ｄＢの範囲内にある、または９９．９％の信頼性で真のノイズ差異の３ｄＢの範囲内にあることができる。
【００９３】
ノイズ推定値のセットが得られたら、最大尤度法によって、格納されたノイズプロフィルと突き合わせることができる。クロストークノイズプロフィルの典型例として、ＩＳＤＮ、ＨＤＳＬ、およびＴ１が挙げられる。
【００９４】
最も適当なプロフィルの選択方法として最尤見積もりを使用するに当たり、得られたノイズサンプルに最も一致するノイズ差異のプロフィルを探す必要がある。プロフィルがどれもそれらしい場合は、数１１を求める必要がある。
【００９５】
【数１１】

【００９６】
ここで、数１２は、ｋ番目のトーン上におけるノイズ測定を含むＬ次元の複素ベクトルであって、Ω_jは、差異を含むｊ番目のクロストークプロフィルである。Ｌは、ｋ番目のトーン上で行われたノイズ測定の番号を示す。この式（数１１）は、数１３のように簡略化される。
【００９７】
【数１２】

【数１３】

【００９８】
ここで、Ｌは、ノイズ差異を推定するために使用されるシンボルの数であり、σ² _j,iはｊ番目のプロフィル内のｉ番目のトーンの差異であり、数１４、数１５であり、Ｎ_k,iは、ｉ番目のプロフィルの時刻ｋにおけるノイズの測定値であり、Ｄ_iはｉ番目のトーン上における経験的なノイズ差異である。上記の式（数１３）によって、ノイズの測定値に最も近いノイズプロフィルを識別することが可能になる。
【００９９】
【数１４】

【数１５】

【０１００】
以下に挙げる３つの実施例では、高速スワッピングおよびビットスワッピングの実現速度を比較する。
【０１０１】
実施例１：１つのビットをスワップする
従来のビットスワッピングで１ビットを移動させる場合は、６０バイトのＡＯＣ帯域幅が必要である。これは、論理的には、双方向のＡＯＣチャンネルを約３０ｍｓ（６０バイト／（２０００バイト／秒））で横断することができる。しかしながら、Ｇ．ライトおよびＧ．ｄｍｔの標準では、（実装の遅延が原因で）８００ｍｓ以内で１ビット以上のスワップ変換を行うことはできない。８００ｍｓは、高速スワッピングには適用されない。逆方向の相互運用性のため、従来のビットスワッピングからこの８００ｍｓの問題を排除することは難しい（しかしながら、高速スワッピングを使用するモデムでは、このような問題は存在しない）。反対に、高速スワッピングは、一方向性のＡＯＣチャンネルのみを使用する８バイトのコマンドを必要とするので、約２１ｍｓ（８バイト／（２０００バイト／秒））＋１７ｍｓ（最悪の場合のタイムアウト）の時間がかかる。
【０１０２】
【表２】

【０１０３】
実施例２：１０トーン上で４０ビットを移動させる
このタイプのスワップとしては、例えばＡＭラジオレシーバのパワーが夜間にレベルアップする場合のような、回線上における緩やかなノイズ変化が代表的である。従来のビットスワッピングでは、論理的に、最低でも６０バイト×（４０ビット）＝２４００バイトが必要であり、１．２秒の時間がかかる。実際にかかる時間は、（１ビットのスワップごとに最低８００ｍｓが必要だと考えて）４０（．８ｓ）＝３２秒である。これに対して高速スワッピングでは、２バイト×（１０トーン）＋４バイト＝２４バイトが必要で、２９ｍｓの時間がかかる（最悪の場合のタイムアウトである１７ｍｓを含む）。したがって、ビット数が多いほどスピードアップも大きくなる。
【０１０４】
【表３】

【０１０５】
実施例３：１００トーン上における４００ビットのプロフィル変更
このタイプのスワップとしては、例えばオフフックのインピーダンス変更等の、回線上における激しいチャンネル変更が代表的である。ビットのスワッピングプロトコルでは、論理的に１２秒そして実際には３２０秒が必要である。この場合は、ビットのスワッピングによって問題を修正している間にエラーを生じる可能性があるので、リテインは明らかに高速であり、使用する必要があると考えられる。高速スワッピングが２０４バイトのＡＯＣコマンドで４００ビット全部を移動させるのに、１１９ｍｓの時間が必要である。ビットスワッピングの高速化は著しく、サービスの介入がないので、リテインより大幅に高速である（ただし、チャンネルの特性が著しく変化した場合は、高速スワッピングインターバルにおいて幾らかのビットエラーを生じる可能性がある）。
【０１０６】
【表４】

【０１０７】
さらに詳述すると、ＥＳの実現には（２ｎ＋４）／Ｒ_AOC＋Ｔ_time-out秒が必要であり、ここで、ｎ＝該当トーンの数、Ｒ_AOC＝バイト／秒で表されたＡＯＣのチャンネル速度、Ｔ_time-out＝タイムアウトの時間である。タイムアウトの時間は、最悪の場合で１７ｍｓであり、もっとずっと速くても良い。タイムアウトの時間とは、ＡＯＣバイトが最後に送信された後、レシーバによって、受信されたチャンネル出力のビットテーブル内において何の変更も見出されなかった場合に、レシーバが、トランスミッタは何らかの理由でＥＳを実現できなかったまたは実現できないと想定するのに必要な時間である。簡略化のため、タイムアウトは、次のスーパーフレーム境界となるように選択される（ただし論理上はもっと小さい）。従来のスワッピングでは、移動されたビット数をｂとした場合に、必要な時間は僅か０．８ｂ秒であった。図７は、ｋ＝１，２，５，１０の各場合に関し、ｂ＝ｋｎの改善度を該当トーン数に対してプロットしたグラフである。（この特定のプロットは、実際には２ｍｓのタイムアウトを使用しており、これはやはり対象範囲内である）図示されるように、ＥＳプロトコルの改善度はいずれの場合も非常に大きい。したがって、このプロトコル自体が、高速スワッピングを行うに足る能力を有する。
【０１０８】
以上からわかるように、本発明は従来技術と比べて多くの利点を有する。例えば、高速スワッピングによってビットスワッピングが高速化される。高速スワッピングの速度がビットスワッピングの１０，０００倍の速度に達する場合もある。さらに、高速スワッピングの実行時間はほとんど瞬間的であり、特に大きいクロストーカが突然導入された場合などに、多数の著しい変更間で接続性を維持するのに充分である。また、高速スワッピングコマンドの使用では、初期化中に追加の情報を引き渡す必要はない。すなわち、確認通知パケットはなく、変更要求コマンドは１度だけ送信されれば良い。さらにまた、高速スワッピングは、より高速な変更が要求される状況における接続の信頼度を高めることによって、再トレーニングの必要性および頻度を低減することができる。
【０１０９】
ビットスワッピングに関連した詳細は、次に挙げる参考文献にも記述されている。すなわち、Salvekar et al.「CHANNEL GAIN CHANGE DETECTION AND CHANNEL PROFILE SELECTION IN A MULTCARRIER SYSTEM（マルチキャリアシステムにおけるチャンネルゲインの変更の検出およびチャンネルプロフィルの選択）」Globecom，１９９９年１２月、John M.Cioffi「VDSL ALLIANCE:VDSL TRANSMISSION SPECIFICATION（ＶＤＳＬ伝送仕様）（００−１１６）」T1E1.4:VDSL，２０００年２月、およびCioffi et al.「G.LITE.BIS:ANALYSIS OF EXPRESS SWAPPING SPEEDS AND RELIABILITY（高速スワッピングの速度および信頼性の分析）」１９９９年４月であり、これらの文献を、本明細書に引用として組み込むものとする。
【０１１０】
次に、スワップの速度を増加させる代替の方法を説明する。この新しい方法は、ともにチャンネルをモニタリングして様々なチャンネル変更のプロフィルを格納するレシーバおよびトランスミッタを含む。処理中に変更が検出されたら、レシーバおよびトランスミッタは、格納された同じプロフィル（１６プロフィルのうちの１つ）に切り換えて動作を継続する。したがって、高速スワップコマンドや他のコマンドがレシーバから送信されることはなく、このためより高速なスワップが行われる。さらに具体的に言うと、レシーバおよびトランスミッタは、チャンネルの性能をモニタリングして信号の適応化を制御するための回線モニタを備えるように構成される。例えば、レシーバおよびトランスミッタが、上述したように回線モニタを備えても良い。レシーバおよびトランスミッタは、また、特定の回線変更に対応する複数の所定プロフィルを格納するようにも構成されている。したがって、モニタによってチャンネル内の変更が検出された場合、レシーバおよびトランスミッタは、関連のデバイスに対し、所望の変更に最も良く一致する所定のプロフィルを選択するように指示する。この方法では、レシーバおよびトランスミッタの両方が、所定のプロフィルで動作することができる。例えば、所定のプロフィルに、マルチトーン信号の各トーンに対するビットおよびゲインの分布が含まれていても良い。
【０１１１】
以上では、幾つかの実施形態を詳細に説明したが、本発明の趣旨および範囲から逸脱しない範囲内で、他の多くの特定の形態で本発明を実現できることを、理解する必要がある。本発明は、特に、離散マルチトーンシステムのための高速スワッピングに関して説明されている。しかしながら、本発明が、他の様々なシステムにおいても使用可能であることを、理解する必要がある。例えば、サブチャンネル（サブキャリアビットおよびゲイン）の動的な適応化を必要とするマルチチャンネルまたはマルチキャリアシステムが挙げられる。さらに、レシーバのモニタリングは様々な方法で実現可能であり、本発明の範囲を逸脱しない範囲内で大幅な変更が可能であると考えられる。したがって、上述した実施例は例示的であって限定的ではないと考えられ、本発明は、ここで挙げた詳細に限定されず、添付した特許請求の範囲内で変更することが可能である。
【図面の簡単な説明】
【図１】 Ｔ１．４１３ビットスワッピング要求のフォーマットを示す図である。
【図２】 典型的なビットおよびゲインテーブルを示す図である。
【図３】 本発明の一実施形態に従って、高速スワッピング要求のフォーマットを示す図である。
【図４】 本発明の一実施形態に従って、モデムのアーキテクチャを示す図である。
【図５】 本発明の一実施形態に従って、レシーバを示す図である。
【図６ａ】 ４ビットアロケーションのトーンに対する典型的な出力コンステレーションを示す図である。
【図６ｂ】 ６ビットアロケーションのトーンに対する典型的な出力コンステレーションを示す図である。
【図７】 高速スワッピングの改善因子対影響されたトーンの数をプロットしたグラフである。
【符号の説明】
１２…メッセージヘッダ
１４…メッセージ
１６…メッセージフィールド
１８…コマンド
２０…トーンインデックス
５０…ビットおよびゲインテーブル
５２…トーンナンバフィールド
５４…ビットアロケーションフィールド
５６…ゲインまたはパワーレベルフィールド
１００…高速スワッピングコマンド
１０２…メッセージヘッダフィールド
１０４…高速スワッピング制御フィールド
１０６…トーンメッセージ
１０８…エラーフィールド
１１０…トーンインデックス
１１２…所望値インジケータ
２００…モデム
２０２…トランスミッタ
２０４…レシーバ
２０６…エンコーダ
２０８…変調器
２１０…ウィンドウ操作フィルタ
２１１…コントローラ
２１２…アナログインタフェース
２１４…アナログインタフェース
２１６…タイムドメインイコライザ
２１８…復調器
２２０…デコーダ
２２１…コントローラ
２２２…リモートデバイス
２２３…伝送メディア
２２４…ビットおよびゲインテーブル
２２６…ＡＤＳＬオーバヘッドチャンネル
２２７…古いビット／ゲインテーブル
２２８…新しいビット／ゲインテーブル
３００…DMTレシーバ
３０２…第１のデコーダ
３０４…第２のデコーダ
３０５…シンドロームジェネレータ
３０６…シンドロームジェネレータ
３０８…スイッチ
４００…４ビットの布置
５００…６ビットの布置[0001]
BACKGROUND OF THE INVENTION
The present invention generally relates to high speed data transmission systems using multi-carrier modulation. More particularly, a fast parameter change command and protocol suitable for use in a multi-carrier transmission system is disclosed.
[0002]
[Prior art]
In recent years, the use of multi-carrier modulation in high-speed modems has attracted attention. For example, the Telecommunications Information Solutions Association (ATIS), a group accredited by the ANSI (American National Standards Institute) Standards Group, is based on discrete multitone for digital data transmission over an asymmetric digital subscriber network (ADSL). Announced the standards. The standard is primarily intended for data transmission over normal telephone lines, but can be used in a variety of other applications as well. The North American standard is called the ANSI T1.413 ADSL standard and is included herein by reference. The transmission rate under the ADSL standard is intended to facilitate the transmission of information at a rate of at least 6 million bits per second (ie 6 Mbit / s) over a twisted pair telephone line. The scaled system defines the use of a discrete multitone (DMT) system with 256 “tones” or “subchannels” each having a width of 4.3125 kHz in the forward (downstream) direction. In the telephone system, the downstream direction is defined as transmission from a telephone office (usually owned by a telephone company) to a remote location that is an end user (ie home or business user).
[0003]
Although the ADSL standard is widely accepted, efforts are being made to improve the T1.413 ADSL standard and to provide ADSL or other standards for communication at other data rates. For example, efforts are currently being made to define a simplified version of the standard that specifically uses only 128 tones. This effort is made by T1.413 and is generally described in G.C. It is called a light standardization initiative. There is also an effort to define criteria for fairly fast data. That effort is called the VDSL (Very High Speed Digital Subscriber Network) standard. The VDSL standard is intended to facilitate a transmission rate of at least 25.96 Mbit / s, preferably at least 51.92 Mbit / s in the downstream direction. In order to realize these speeds, it is generally necessary to make the transmission distance on the twisted pair telephone line shorter than the length allowed by ADSL. At the same time, the Digital Audio Video Association (DAVIC) is working on similar systems. The system is called Fiber To The Curve (FTTC). The transmission medium from the “curve” to the customer's building is a standard unshielded twisted pair (UTP) telephone line.
[0004]
One problem inherent in high speed DSL modems using multi-carrier modulation is how to handle line state variations. For example, in the T1.413 standard and other proposed DMT-based systems, the communication modem undergoes a short training period before data communication begins. During the training period, test signals are transmitted to effectively test line quality at various frequencies. In general, line quality is determined by the signal / noise ratio (SNR) at each tone. Next, the number of “bits” assigned to each tone is determined primarily based on the detected training signal. However, after a training period, changes may occur on the transmission line that may affect the ability to transmit information at the rate assigned in several tones. The change in the transmission line is caused by the following various causes. The change of the line caused by the temperature of the handset being raised or lowered by the customer, the change of the crosstalk noise due to the activation or inactivity of the adjacent line, the increase of the AM radio signal at night.
[0005]
As time goes on and the quality of the line degrades, errors are more likely to occur and some work needs to be done to adjust the allocated bit allocation. One way to adjust bit allocation is to simply retrain the modem. However, retraining has the disadvantage that it takes a relatively long time, resulting in a short outage. Another method of adjusting the bit allocation defined by the T1.413 standard is a procedure called “bit swapping”. The bit swapping protocol is intended to reduce the amount of information transmitted on a tone by a certain number of bits when an error is detected on the tone. If other tones appear to have extra SNR, the amount of information transmission is increased by a corresponding amount.
[0006]
In the T1.413 standard, the bit swapping protocol is clearly defined. More specifically, if a receiver determines that bit swapping is required, the receiver sends a request for bit swapping through an overhead channel (commonly referred to as an AOC-ADSL overhead channel). The bit swapping request has the specified format shown in FIG. As shown, the first byte of the bit swapping request is the message header 12. The message header 12 consists of all headers that identify the command as bit swapping. The message header 12 is followed by an 8 (or 12) byte message 14. Message 14 is divided into 4 (or 6) segments, each segment being referred to as a message field 16. Each message field 16 includes a 1-byte command 18 followed by a 1-byte tone index 20. The tone command identifies the tone to which the command applies. The 1-byte command includes the following functions. Add bit, erase bit, increase 1, 2, 3 dB power, decrease 1, 2, 3 dB power, do nothing, and unique commands.
[0007]
T1.413 further requires that a 3 byte bit swapping response command be sent back to the unit that requested the bit swapping to confirm receipt of the bit swapping request. The bit swapping response command specifies a particular symbol count for which swapping is performed. Although the response command is simply used to detect the implementation of the new bit distribution, the response slows down the swapping and can still cause a failure if no response is received.
[0008]
The T1.413 protocol further states that the bit swapping request command is transmitted five times in succession and that only the receiving unit responds to the bit swapping command when the receiving unit receives most of the five transmissions. Order. Therefore, 45 bytes are required to request the movement of 1 byte and 15 bytes to respond to the request (and it must be repeated 5 times). If the waiting time is ignored, the shortest time for swapping is on the order of 30 ms (60 bytes at 16 kilobytes / second). However, current standards actually require that swapping does not occur more often than once every 800 ms. This allows simplification of the transceiver, but also slows the swapping process if more than four tones are changed. Therefore, standard ADSL bit swapping allows for slow variations in the transmitter. However, with the advent of splitterless ADSL and the general spread of DSL, it has become clear that DSL lines are subject to sudden changes that require significant changes in bit distribution. A more efficient mechanism is needed to reduce the time to perform bit redistribution in a multicarrier transmission system, taking into account the disadvantages of slow and retrained standardized bit-swapping protocols It became clear that.
[0009]
SUMMARY OF THE INVENTION
To achieve the following and other objectives of the present invention, a method and device for adaptively changing communication signal parameters (such as subcarrier gain or bit allocation) in a multicarrier based transmission system is disclosed. ing. In one aspect of the invention, the unit that determines the need for change sends a change request to the second unit. The change request identifies one or more specific subcarriers to be changed and the desired values for the parameters for each identified subcarrier. The requesting unit then monitors the received communication signal to determine whether the requested change has been realized. The determination of whether the requested change has been realized is based at least in part on an analysis of a portion of the received communication signal that was to be changed.
[0010]
In a preferred embodiment, the change request is suitable for identifying a plurality of specific subcarriers to be changed, as well as a desired value for a parameter for each identified subcarrier. For example, the parameter may be a desired bit allocation or a desired power level (gain) for the associated subcarrier. The desired value may be an absolute value (eg, transmitting 8 bits on this tone) or a relative value (eg, increasing the number of bits transmitted on this tone by 2). In other preferred embodiments, the protocol does not include an explicit response that a change request has been received or realized.
[0011]
Monitoring may be done in various ways. For example, in some embodiments, the request unit redundantly decodes the received communication signal using both the current value of the parameter and the desired value of the parameter. In this embodiment, the determination of whether the requested change has been realized is made based at least in part on the decoding. For example, the determination of whether the requested change has been achieved may be based at least in part on the analysis of errors detected using the respective decoding and the selection of a decoding with fewer errors. Another approach is to generate a first forward error correction syndrome based on the decoding of the current value and a second forward error correction syndrome based on the decoding of the desired value. In this approach, the determination of whether the change request has been realized is based on an analysis of its syndrome.
[0012]
In some embodiments, the monitoring step includes monitoring one or more specific subcarriers that are to be changed by the change request. If a change is detected on a particular subcarrier, it is determined that the change has been realized. Again, monitoring may be done using various mechanisms. For example, if the multicarrier signal is a DMT signal, the energy level of one or more tones can be monitored. For example, a change may be made that increases the allowable power for that tone. In this scenario, it can be determined that the requested change has been realized if more energy than expected with the current parameter value is detected on that tone. Alternatively, one of the tones can be zeroed or a tone that has been zeroed can be activated. These types of approaches are relatively easy to implement and make it easy to detect the realization of changes without requiring explicit feedback from the unit requesting the change.
[0013]
In another aspect of the invention, the change request command includes a header, fast swap control, at least one subcarrier identifier, at least one desired parameter value indicator, and an error field. The header identifies the command as a change request command. Fast swapping control specifies a modified tone count. The tone count indicates the number of tones changed by the change request command. Each subcarrier identifier identifies a specific subcarrier to be changed by a change request command. Each desired parameter value indicator identifies a desired value for the parameter of the associated subcarrier. The error field allows the unit receiving the change request to detect whether there is an error in the interpretation of the change request command. In some embodiments, the superframe number may be included in the header. In other embodiments, the super frame number may be included in the fast swap control. Furthermore, the superframe number specifies how many superframes follow, or in which superframe high-speed swapping occurs.
[0014]
In a preferred embodiment, the desired parameter value indicator identifies at least one of a desired bit allocation and a desired gain for the associated subcarrier. In some embodiments, one byte can be used to identify both the desired bit allocation and the desired gain for a particular subcarrier.
[0015]
In another aspect of the invention, an improved modem design is disclosed that includes a duplicate decoder in the receiver. The duplicate decoder is configured to decode the demodulated multicarrier signal using various parameters for at least one subcarrier that is to be changed by the change request command. An analyzer is then provided to determine which duplicate decoder has decoded the correction signal.
[0016]
In one embodiment, the overlapping decoder is a subcarrier decoder configured to decode the same subcarrier. In other embodiments, the overlap decoder is a signal decoder configured to decode a plurality of subcarriers of a multicarrier signal. In one particular implementation, the analyzer includes a duplicate syndrome generator configured to generate a syndrome of the signal decoded by the duplicate decoder.
[0017]
DETAILED DESCRIPTION OF THE INVENTION
With reference to some preferred embodiments, the relevant figures are used to describe in detail methods, devices, and protocols suitable for implementing high-speed bit swapping in a multi-carrier transmission system. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention. However, it will be apparent to those skilled in the art that the present invention may be practiced without some or all of these specific details. Well-known process steps have not been described in order not to unnecessarily obscure the present invention.
[0018]
In multi-carrier based communication systems, it is generally desirable to facilitate the process of adaptively changing channels. In some current implementations (such as the T1.413 ADSL standard), the concept of bit swapping is used to facilitate the process of adaptively changing channels. However, the standard bit-swapping configuration is relatively slow and therefore not particularly effective when dealing with many channel changes. Therefore, the present invention intends to provide an improved mechanism for realizing adaptive changes to the channel.
[0019]
In general, the present invention contemplates that when the transceiver determines that the change is valid, the transceiver sends a fast change request to the other party. The requesting transceiver then monitors the signal that determines whether the requested change has been realized and decodes the received signal accordingly. For purposes of illustration, the present invention has been described using a DSL-based point-to-point bi-directional communication system using discrete multitone modulation. However, it will be appreciated that the described technique is readily applicable to any multi-carrier transmission system, not just point-to-point systems.
[0020]
In both conventional bit swapping (SB) and the disclosed constrained swapping (ES), the receiver determines the need for change based on the amount of error occurring in the input signal. If it is determined that a change is necessary, the receiver determines the best possible bit allocation for the problematic tone. For example, if the receiver measures an increase in noise (noise sufficient to cause an error) on a particular tone, the receiver reassigns a portion of the bit from that tone to another less noisy tone . The receiver stores the desired bit allocation in a table (commonly referred to as a bit and gain allocation table). In most cases, the bit and gain allocation table has entries for all the tones used in the multi-carrier signal, and each tone entry includes bit allocation and gain or power level for the corresponding tone. When a change is desired, the receiver sends a request to the transmitter requesting a change in the signal corresponding to the new allocation.
[0021]
For ease of discussion, FIG. 2 generally shows a typical bit and gain table 50. The bit and gain table 50 generally includes a tone number field 52, a bit allocation field 54, and a gain or power level field 56. The tone number field 52 includes G. For operation in accordance with the light standard, an entry for 128 tones, For operation compliant with the dmt standard, it contains entries for 256 tones. In other systems (such as the VDSL plan), the number of entries can vary widely since quite a lot of tones have been proposed. In addition, each tone entry (one of 128 or 256 tones) typically includes a bit allocation that determines the number of bits transmitted on the associated tone. The bit allocation for each tone is typically included in the bit allocation field 54. The number of bits assigned to a particular tone varies widely based on the line conditions, the desired transmission rate, the protocol used, etc. For example, the number of bits assigned to each tone is G. In the light transmission system, G. In the dmt transmission system, it varies in the range of 0-15. Also, these numbers can vary widely in other systems. In addition, each tone entry generally includes a gain allocation that determines the transmitted energy level with the associated tone. The gain allocation for each tone is typically included in the gain allocation field 56.
[0022]
Next, referring to FIG. 3, the format of the high-speed swapping command will be described according to an embodiment of the present invention. As described above, if the receiver determines that a change is necessary, the receiver sends a fast swap command to the transmitter. The fast swap command 100 includes a message header field 102, a fast swap control field 104, a series of one or more tone messages 106, and an error field 108. The header field 102 includes a header value that identifies the command as a change request command. The fast swap control 104 generally includes the number of superframes that follow, or a superframe number that indicates the superframe on which fast swapping is performed, and a tone count that indicates the number of tones that are changed by the change request command. Each tone message 106 includes a tone index that identifies the subcarrier to be changed by the change request command, and a desired value indicator that indicates the desired bit allocation and gain parameters for the identified tone. The error field 108 allows the unit that receives the change request to detect whether there is an error in the interpretation of the change request command.
[0023]
More specifically, the message header 102 identifies that the command is a change request command and is of an appropriate size. For example, in the described embodiment, the message header occupies one byte and includes a bit pattern (eg, 11110011) that is predefined as a fast swap command. G. dmt or G.M. In a write implementation, the swapping command is transmitted on the auxiliary overhead channel (AOC), and any message transmitted with an AOC starting with a defined pattern is immediately recognized as a fast swapping command. Furthermore, although the header is shown before the command, it should be noted that the position of the header is not limited and may be located anywhere in the command.
[0024]
The message header 102 is followed by a fast swap control 104. It indicates the superframe to be fast swapped and the number of tones changed by the change request command. For example, if the tone (# 2, # 80, # 95) is to be changed by a change request, the tone count is 3. In the described embodiment, the fast swap control 104 also occupies one byte. For example, the fast swap control 104 sets the most significant bit to 0 if the transmitter performs fast swapping in the next superframe, and if the transmitter performs fast swapping in the next superframe. Is set to 1. In addition, the remaining bits are lined with the number of tones n, which are changed by the next 2n bytes in the command. G. In the light standard, 128 tones are used, so one byte (remaining 7 bits) is sufficient to accommodate any number or all of the tones modified with one command. However, G. Since 256 tones are used in the dmt standard, it is desirable to increase the size of the fast swap control (eg, to 8 bits). In addition, other systems have considered using significantly more tones than 256. For example, in the proposed VDSL system, it is planned to use 4096 tones. Even in such a system, it is desirable to increase the size of the fast swapping control (eg, to 12 bits).
[0025]
It should be noted that although the superframe number is described as part of the fast swapping control, this is not a limitation and can indicate the timing of the change in various ways. For example, the superframe number or superframe designation may be part of the message header, and the timing may be part of a fixed agreement between the receiver and transmitter. For example, the super frame number can be specified using various message headers. Furthermore, it should be noted that the superframe number is not necessarily required. If not needed, the superframe number can be omitted from the command.
[0026]
The fast swap control 104 is followed by a series of tone messages 106 each including a tone index 110 and a desired value indicator 112. Each tone index identifies a particular subcarrier that is changed by a change request command. Each desired value indicator indicates a desired value for the tone parameter that is changed for the tone identified by the associated tone index. In the disclosed embodiment, tone bit allocation and tone power (gain) are two parameters that are modified. Therefore, desired value indicator 112 identifies the tone bit assignment and gain for the tone identified by the associated tone index. The desired value is also in the form of an absolute value (eg, transmitting 8 bits on this tone and setting the gain to +1.5), but also increasing the relative value (eg, the number of bits transmitted on this tone by 2). The gain may be increased by a certain level).
[0027]
In the described embodiment, the tone index 110 and the desired value indicator 112 each occupy 1 byte and again can be resized according to the needs of the system. The size of the tone index depends mainly on the number of available tones. If more than 256 tones are available, the tone count needs to be greater than 1 byte. However, G. dmt standard (assuming 256 tones) and G.D. For the write standard (assuming 128 tones), one byte is sufficient. For example, in the proposed VDSL system, 4096 tones are assumed, so for such an implementation it is desirable to increase the size of the tone index to 12 bits.
[0028]
Desired value indicator 112 is associated with the upper nibble of desired value indicator 112 (ie, the first 4 bits) indicating the desired tone bit assignment and desired gain for the tone identified by associated tone index 110. Indicates the new number of bits assigned to the tone. The maximum number of bits for any tone is G. 11 for light (15 for G.dmt). Therefore, the upper nibble is 0000 for 0 bits, 0010,. . . , 15 bits may be encoded as 1111. The lower nibble of the desired value indicator 112 (ie, the last 4 bits) indicates the desired gain level. With the 4 bits assigned, 16 different gain levels can be specified. Again, the desired gain value may be in the form of an absolute level or in the form of a relative value change (eg, power increased by 1, 2 or 3 dB, power decreased by 1 or 2 dB, unchanged). . For example, in the discrete multitone example, 16 different gain levels from -4 to +3.5 (every 0.5 dB) for several nominal gain levels, with each potential for a particular gain level. Can be specified with any value. Although a uniform gain level has been described, it should be understood that there are various gain level distribution methods that can be used to change the desired gain. Furthermore, not only the relative size of the increase, but also the number of gain levels can vary widely depending on the needs of a particular system.
[0029]
Also, the amount of increase need not be uniform. If more than 15 bits can be transmitted on any tone, or if a power level of 16 or more is desired, it is preferable to provide 1 byte for the desired value indicator 112. For example, increasing the size to 2 bytes will substantially increase the resolution. Of course, if different, additional or fewer parameters are controlled for a particular subcarrier or tone, the desired value indicator can be adapted to carry the appropriate information accordingly. .
[0030]
In general, if the two modems are not synchronized, the next swapping command will clearly indicate the desired value, so using an absolute value as the desired value will change (relatively) by increasing the bit and gain allocation. It is considered somewhat more reliable. Conversely, when changing by increasing parameters, such as bit swapping, there is no mechanism to resynchronize the lack of retraining of the system if the system goes out of sync.
[0031]
Since the fast swap command can contain a bit error, the error field 108 is used to detect the error. The error field contains a value (eg, checksum) used to verify the interpretation of the fast swap command by the transmitter. If the transmitter detects an error in the interpretation of the fast swap command, the transmitter does not implement the requested change. Error detection is used so that the transmitter can only implement change requests that are effectively transmitted without errors. Also, the size of the error field 108 can vary widely to facilitate the implementation of an appropriate error checking configuration. For example, in the illustrated embodiment, the error field 108 occupies two bytes after the fast swap command and stores a CRC (Cyclic Redundancy Check) value.
[0032]
CRC is well known to those skilled in the art and will not be described in detail for simplicity. However, if there is a 13 dB margin loss in the transmission of the high-speed swapping command (the reverse link may be temporarily affected by the same failure as the forward link, which may temporarily show a high error probability) Even so, it should be noted that CRC failures are estimated to occur only approximately once every 20 years. Therefore, high-speed swapping command transmission is really reliable to prevent catastrophic events that require line repair.
[0033]
In another embodiment of the invention, the fast swap command may be modified slightly to accommodate other formats such as VDSL. In one such embodiment, in detecting a sub-channel SNR change, the receiver initiates fast swapping by sending a fast swap request back to the transmitter via the VDSL overhead channel (VOC). Again, the high-speed swapping command is transmitted only once, and the bit distribution (or gain distribution) of the n tone can be changed by transmitting the command. In one implementation, the VDSL fast swap command includes a VOC message field, a fast swap control field, a series of one or more tone message fields, a dummy bit field, and an error field. The VOC message header occupies 1 byte and consists of either 11110010 or 11110011 patterns and indicates a guaranteed fast swapping request. The header pattern 11110010 means that fast swapping should be performed on the next bit swapping frame, and the pattern 11110011 means that fast swapping should be performed on the next next bit swapping frame. . Fast swapping control typically uses 12 bits and is configured to indicate the total number (n) of tones whose bit / gain distribution needs to be updated. Alternatively, the VDSL command includes one header (11110011), one bit for designating the number of bit swapping frames for performing high-speed swapping or the number of frames for performing high-speed bit swapping, and 12 bits for designating the tone count. By having only the ES control including, it may be configured similarly to the above-described high-speed swapping command.
[0034]
A tone message typically uses 20 bits and includes a subchannel index and a desired value indicator. The first 12 bits indicate the subchannel index, and the next 8 bits indicate the desired value indicator. The 4-bit upper nibble of the desired value indicator encodes a new absolute number of bits (0 bits is 0000, 2 bits are 0010, 15 bits is 1111) and the 4-bit lower nibble is Encode the desired gain level for some known nominal value as the most important bit sign bit (increment by 0.5 dB) 2's complement 4 bit quantity between -4 and +3.5 by). The dummy bit field is configured to adjust the total bit amount for the fast swap command so that the command is uniformly written in bytes. Therefore, if n (the number of tones that need to be updated) is an even number, the dummy bit field adds 4 bits to the command, and if n is an odd number, the dummy field Add a bit. The error field typically uses 16-bit CRC protection for error detection. Furthermore, it should be noted that dummy fields are not a limitation and commands may be configured in other ways. For example, the command may be configured to have a variable length error field.
[0035]
Although specific bit counts and specific bit patterns for certain commands have been described in the implementation examples, the actual number of bits allocated and the specific bit pattern used for a specific command or field is not necessary in any case. Obviously, it can be varied widely to accommodate.
[0036]
In bit swapping, response commands are typically sent by the transmitter to the receiver to adjust the time at which the swapping is performed. That is, the transmitter sends information to the receiver that the change is achieved with a particular symbol. In contrast, in fast swapping, the receiver that initiates fast swapping monitors the returned signal and determines whether the command has been implemented by the transmitter. The described fast swapping includes a 0-bit swapping response command. Since these response commands increase the execution time for swapping and are preferably avoided in fast swapping, they require a receiver to know the time when fast swapping took place.
[0037]
Referring now to FIG. 4, a modem architecture suitable for implementing the disclosed high speed swapping transmission scheme will be described in accordance with one embodiment of the present invention. Modem 200 includes a transmitter 202. The transmitter incorporates several elements including an encoder 206, a discrete multitone modulator 208, a windowing filter 210, and a controller 211. The encoder 206 has a function of multiplexing, synchronizing, and encoding transmitted data (video data or the like). More specifically, the encoder converts input bits into in-phase quadrature elements for each of the multiple subchannels. The encoding can be performed using various error correction methods. For example, forward error correction works well. Encoder 206 is typically configured to output a number of subsymbol sequences equal to the number of subchannels available to the system. For example, in a system having 256 subchannels, the encoder 206 outputs a value obtained by subtracting the number of subchannels from a 256 subsymbol sequence in a limited frequency band. These inputs are composite inputs that are passed through discrete multitone modulator 208. Modulator 208 is generally an IFFT modulator that calculates the inverse Fourier transform by any suitable algorithm.
[0038]
After the encoded signal is modulated to form a discrete multitone signal, the modulated signal is passed through windowing filter 210 and / or other filters to minimize the lack of band energy. This is desirable to prevent saturation of the remote receiver analog interface. The window operation can be performed by various conventional window operation protocols. The transmitter also includes an analog interface 212. The analog interface sends a discrete multitone signal to the transmission medium. In a wiring system such as a twisted pair telephone line or coaxial cable, the analog interface 212 may take the form of a line driver.
[0039]
Modem 200 also includes a receiver 204 for receiving a multitone signal from the transmitter. The receiver 204 generally includes an analog interface 214, a time domain equalizer (TEQ) 216, a demodulator 218, a decoder 220, and a controller 221. The signal received by modem 204 (from the transmitter) is first received through analog interface 214. The time domain equalizer 216 effectively performs a filtering function on the received signal. A window operation filter (not shown) may be used. The demodulator 218 demodulates the equalized discrete multitone signal, and the decoder 220 decodes the demodulated signal. Demodulator 218 and decoder 220 effectively perform the opposite functions of modulator 208 and encoder 206, respectively. For example, demodulator 218 is typically an FFT modulator that calculates a Fourier transform by any suitable algorithm. The decoded signal is then passed from the decoder 220 to a remote device 222, such as a videophone, television, computer, or other suitable receiving device.
[0040]
In a discrete multitone system (DMT), the bit distribution of the DMT signal is adaptively determined so as to improve the transmission performance of the system. To facilitate this, the system typically includes a line monitor (not shown) that monitors the communication line and determines the quality of each available subchannel. In one embodiment, the line monitor determines noise level, gain and phase variation for each subchannel. The purpose is to evaluate the signal / noise ratio of each subchannel. Therefore, other parameters may be monitored in addition to or instead of the above parameters. The determination of the amount of data transmitted in each subchannel, as well as the determination of the subchannel for transmitting encoded data, is made dynamically based on several factors. The elements include detected channel quality parameters, subchannel gain parameters, allowable power mask, and desired maximum subcarrier bit error rate. It is noted that these factors need not be constant between subchannels and may change during use. More noteworthy, the line quality parameters can be examined repeatedly, adjusted in the modulation scheme in real time, as the line quality of the various subchannels changes during use. Dynamically adjust the modulation. Channel monitoring is described in detail below.
[0041]
In most configurations, the receiver 204 is configured to monitor channel performance and control DMT adaptation. For example, the controller 221 may include a line monitor as described above. In particular, the receiver 204 monitors the channel output signal to confirm the need for changes to the transmitter bits and gain table 224. The table stores the number of bits and the corresponding transmission energy (or equivalent) and is used by each DMT tone (or subchannel). Both transmitter 202 and receiver 204 store copies of the same table, each using them for encoding and decoding. For example, the table shown and described in FIG. 2 may be used. As will be appreciated, changes in these tables that cause an improvement in transmission performance as transmission lines change over time are determined by continuous monitoring. In addition, an AOC (ADSL Overhead Channel) 226 is typically returned from the receiver 204 to the transmitter 202, as indicated by the dotted line. As is well known, AOC 226 is a dedicated portion of the full bandwidth of the system used between the receiver and transmitter for communication. For example, the first 32 tones of a 256 tone system may be reserved for AOC 226.
[0042]
In determining the need for change, a command (eg, a request for change) is sent to the transmitter 202 through the AOC 226. In most embodiments, a command is used to instruct the transmitter 202 to change the configuration of the transmitter bit and gain table 224 (eg, the number of bits and / or transmission energy level of each tone used for DMT transmission). Used. For example, the command may be a fast swapping command shown and described in FIG. After receiving the command, the transmitter 202 implements the requested change. As described above, error detection may be used on the AOC 226 so that the transmitter 202 can only implement change requests that have been successfully transferred without errors.
[0043]
In fast swapping, the receiver 204 is configured to monitor the incoming DMT signal to determine if the requested change has been made by the transmitter. More specifically, the receiver 204 is configured to recognize a tone (subchannel) for a currently used signal (“current signal”) and a tone for a desired signal (“request signal”). With regard to this information, the receiver can confirm whether the change request has been realized by comparing the tone of the input signal with the tone of the current signal or the tone of the desired signal (change request signal). Taking the current signal as an example, if the tone of the input signal is substantially the same as the tone of the current signal, the receiver can determine that the request for change has not been realized. On the other hand, if the tone of the input signal is substantially different from the tone of the current signal, the receiver can determine that the request has been realized. It should be noted that the present invention is not limited to using the current signal or the change request signal separately, and may be used in combination. That is, the tone of the input signal can be compared with both the tone of the current signal and the tone of the change request signal.
[0044]
More specifically, high speed swapping receivers typically include an old bit / gain table 227 and a new bit / gain table 228. The old bit / gain table 227 relates to the signal that was used before the receiver 204 determined the need for change, and the new bit / gain table 228 was the signal requested after the receiver 204 determined the need for change. It is about. The decoder 220 uses both tables to determine whether the input signal is old or new. That is, the input signal can be decoded according to the old channel gain (old table), the new channel gain (new table), or a combination of both the old and new channel gains (old table and new table). Accordingly, if the input signal operates with the old bit / gain table 227, the decoder 220 will have fewer errors when comparing the signal to the old bit / gain table 227, and the decoder 220 will have the signal replaced with the new bit / gain table. The error when compared to 228 will tend to be numerous. Conversely, if the input signal operates with the new bit / gain table 228, the decoder 220 will have fewer errors when comparing the signal to the new bit / gain table 228, and the decoder 220 will not be able to compare the signal with the old bit / gain table 227. There will tend to be a lot of errors when comparing with. Accordingly, the receiver 204 can determine whether changes have been made based on these decodes.
[0045]
Furthermore, even if the receiver can make the above comparison, it needs to be done relatively quickly. Therefore, the receiver is configured to analyze the input signal for a certain period of time, and if no change request is detected within this period, the receiver recognizes that the transmitter has not realized the change. To do. For example, after the last AOC byte has been sent, if the receiver recognizes that there is no change in the bit table of the received channel output, the receiver could not perform fast swapping for some reason, or Recognize that you can't do it. In one embodiment, if a change request is not detected while combining the downstream and upstream latency and interruption times, the receiver recognizes that the transmitter did not implement the command. Next, if the performance is unacceptable, the receiver may choose to retransmit fast swapping commands, use other correct commands, or perform retraining. In one particular embodiment, the interruption time is chosen as the next superframe boundary. It seems to work well. In this embodiment, a worst case interruption time of 17 ms is provided. However, the interruption time is configured to be very fast. For example, the interruption time may be configured to be about 2 ms.
[0046]
Having described the modem architecture in detail, it should be understood that the modem architecture can be implemented in many other specific forms without departing from the spirit and scope of the present invention.
[0047]
In accordance with one aspect of the invention, a request receiver is configured to use two decoders to decode an input signal with an old bit / gain table (eg, current signal) and a new bit / gain table (eg, change request signal). Has been. The determination of whether the requested change has been realized is made based at least in part on these two decodes. For example, the determination of whether or not a change request has been realized can be made at least in part based on analysis of errors detected using each decode and selection of a decode with the least amount of errors. Typically, the signal with the least amount of error is the signal being transmitted. Thus, the receiver can determine whether the change request has been realized.
[0048]
With reference to FIG. 5, a dual decoding receiver 300 will be described in accordance with one embodiment of the present invention. The dual decoding receiver 300 includes a first decoder 302 and a second decoder 304 that is substantially similar to the first decoder 302. In this embodiment, the first decoder 302 is configured to store a bit / gain table 227 associated with a first set of tones (eg, the current signal), and the second decoder 304 It is configured to store a bit / gain table 228 associated with a set of tones (eg, a new signal that includes a request for change).
[0049]
The dual decoding receiver 300 also comprises a first syndrome generator 305 associated with the first decoder 302 and a second syndrome generator 306 associated with the second decoder 304. For example, an FEC decoder using forward error signal correction (FEC) syndrome may be used. The first syndrome generator 305 is configured to use the current channel gain and noise to calculate a first syndrome for the received tone of the input signal. Similarly, the second syndrome generator 306 is configured to use the new or estimated channel gain and noise to calculate a second syndrome for the received tone of the input signal. A correct match can be obtained by a syndrome generator that detects a minimum amount of errors. Here, it should be noted that the receiver 300 is generally configured to calculate the syndrome when it knows that a swap will occur (eg, after a change request). The dual decoding receiver 300 also includes a switch 308 that is generally used to release a signal having a collect syndrome. The controller 221 is usually used to control a switch 308 provided in the controller 221. However, this is not essential (as shown).
[0050]
More specifically, in the calculated syndrome, the collect match is generally indicated by 0, and the collect match is generally indicated by 1. For example, if the calculation result of the first syndrome generator 305 is 0 and the calculation result of the second syndrome generator 306 is 1, the input signal correctly matches the bit / gain table of the first tone set. The receiver 300 recognizes that there are still change requests to be realized. As described above, when this occurs over a specific period, the receiver times out and recognizes that the change request has not been fulfilled. In this case, the receiver may choose to send another fast swap command, some other command, or retrain if performance is unacceptable. Conversely, if the calculation result of the first syndrome generator 305 is 1 and the calculation result of the second syndrome generator 306 is 0, the input signal correctly matches the bit / gain table of the second tone set. The receiver knows that the change request has been realized.
[0051]
Further, if the calculation results of these syndromes are not 0, it is considered that the line state was not the correct high-speed swap originally requested. This is a receiver design error and the line has become inoperable for the allowed bit distribution, so the modem needs to retrain or activate the indicator (eg, bad LED). That is, if the receiver system is unable to repair the line due to fast swap, the maintenance state of the line needs to be retrained or signaled. Furthermore, when the calculation results of these syndromes are both 0, it is considered that the error detection fails because the error pattern of channel bits / bytes corresponds to a different (incorrect) code word. The probability that such a failure will occur is much less than the probability that a CRC failure will occur, so it can be assumed to be zero in practical applications. If such a failure nevertheless occurs, the receiver can choose to delay the output further by checking the next codeword and then picking the current table and the previous table. The failure rate is usually so low that no further reduction is necessary. Here, even this unusual type of failure is not destructive or weak to the modem, it just means that the modem will continue to work well if an error (one codeword) occurs. It is necessary to be careful.
[0052]
In accordance with another aspect of the invention, determining whether a change request has been realized includes monitoring one or more tones that are intended to be changed by the change request. If a change is detected on these tones, it is assumed that the change has been realized. Such a tone is more reliable than the others because fast swap does not require any change otherwise. The energy level and bit allocation of one or more tones are monitored. For example, the change increases the power available for a particular tone. In this scenario, it can be assumed that the required change has been realized if more energy is detected on that particular tone than if the current signal was used. In another embodiment, the change reduces the number of bits for a particular tone. If less energy is detected on that particular tone than using the current signal, it can be assumed that the requested change has been realized.
[0053]
In one embodiment, the requesting receiver is configured to monitor at least one tone of the modulated output (FFT output). These tones are preferably tones for which an increase in the number of bits is required due to high-speed swapping. These tones are generally more robust (eg, less noisy) as they can handle a larger number of bits. However, it should be noted that this is not limiting and that a reduced number of bits may be used. The determination of whether the change request has been realized is based at least in part on a change in the output configuration of the input signal. The output configuration illustrates the distribution of bits in a specific tone. The output configuration generally includes both the phase shift and the amplitude of a particular tone. To facilitate discussion, FIG. 6a shows a representative FEQ (frequency equalizer) output configuration 400 for tones assigned 4 bits. Since the output configuration is well known in the art, further description is omitted for simplicity.
[0054]
More specifically, whether the request for change has been realized by comparing the first output configuration of a particular tone in the first signal with the second output configuration of the same tone in the second signal. It is determined. If the second output configuration is greater than the first output configuration, the receiver knows that the change request has been realized. For example, if the number of bits on a particular tone increases by 2 due to a change request, the point on the output configuration for that tone will increase fourfold. To facilitate discussion, FIG. 6b shows a 6-bit output configuration 500 (ie, a 4-bit output configuration 400 after 2 bits have been increased). There are 16 points in the 4-bit output configuration 400 and 64 points in the 6-bit output configuration 500. Assuming that FIG. 6a represents the first signal and FIG. 6b represents the second signal, the receiver can determine whether the change request has been realized based on the nature of the received signal. For example, the maximum gain (amplitude) of the first signal (obtained from a 16 point configuration) is less than the maximum gain (amplitude) of the second signal (with 64 configuration).
[0055]
In certain implementations, output configuration comparisons may be made by utilizing a pair of overlapping tone decoders for a particular tone. The decoder typically has a plurality of individual tone decoders for each tone used in the multitone signal (eg, 128 for G. light, 256 for G. dmt). . Each tone decoder outputs a specific output configuration for that tone when it decodes the transmitted signal. When the number of tone decoders for a particular tone is doubled, the change request is made by comparing the first transmission signal with the first output configuration with the second transmission signal with the second output configuration. Whether or not is realized. However, although it is possible to double the number of specific tone decoders, only a few tone decoders need to be doubled to determine the change with appropriate reliability.
[0056]
Importantly, in the presence of noise, a gain change of 1 dB or 2 dB cannot be correctly identified. However, gain changes on one tone that occur in connection with gain changes on other tones can be detected with higher confidence. Thus, the receiver is not limited to monitoring one tone in determining whether a change request has been realized. The receiver may monitor multiple tones that have resulted in bit increases and / or decreases and additional gain changes.
[0057]
In another embodiment, the receiver may be configured to temporarily zero out the good tone in the fast swap command and determine whether the change request has been fulfilled by the transmitter. The lack of transmitted energy on that tone indicates that a change request has been fulfilled by the transmitter. A second fast swapping command that later re-activates the same tone (and is rediscovered by the presence of energy) can restore the margin to full level. The effect of one tone loss on performance is usually small, but generally it also depends on the channel.
[0058]
From the above, it can be understood that accurate and rapid channel identification is very important in order to effectively use high-speed swap (or any swapping) in the DMT system. Inaccurate or inaccurate channel monitoring is much more prone to DMT transmission failures than any other reasonable line change, but this happens suddenly. Thus, several embodiments for successfully monitoring a channel are described. Here, it is necessary to understand that the method described below is not limited, and there are many original mechanisms for monitoring the state of the line. It should also be understood that these methods are generally most useful in dealing with channel gross changes rather than minor differences in the bit / gain distribution. For example, a channel gross change may be so gross that it causes a slicer bit error in at least one tone of the DMT system. If such channel changes can be identified, fast swapping is useful to quickly correct the channel in this type of situation.
[0059]
For example, gross changes are strongly indicated by at least three types of measurements. That is, the syndrome is non-zero, the CRC violation is at the superframe level, and the instantaneous or short-term mean square error / noise estimate is high. The receiver can measure any or all of these three in a trivial manner and decide to enter the gross change state. In the gross change state, the output power of the channel can be measured quickly. Since noise is typically at least 10 dB below the signal transmitted to the receiver on any tone of interest, the instantaneous channel output power estimate is accurate enough to confirm channel gain changes. Have For example, channel gain changes caused by off / on-hook impedance changes can affect many subchannels at the same time, and can be easily distinguished from newly activated crosstalk noise.
[0060]
The channel change generally corresponds to a change in the measured channel transport function. These changes are as prominent as changes such as line attenuation and channel delay. Channel temperature changes cause channel changes that are very slow and can be fully handled by either SB or ES. Changes in humidity caused by wet cables during rainy weather are rapid (and often cause circuit failure). If the humidity is low enough and the channel can remain operational, the channel can change rapidly in 1 second or less. Changing the off / on-hook impedance is the fastest type of channel gain change and typically causes an attenuation loss / gain of 1-10 dB across almost all transmission bands.
[0061]
To detect gain changes, the receiver monitors the total received power. When the received power changes greatly, the gain path is tracked and it is estimated that the noise power has not changed. Changes in the total output power of the channel are generally detected by monitoring the sum of the squares of the channel output samples. A 1 dB or more change in channel output power can be estimated with high confidence (1 ms for G. light sampling rate, even faster for other DSLs) by using 1000 or fewer samples. Can do. A way to identify the channel change dependency on frequency (tone) is to divide the channel FFT output by the known sync symbol FFT the next time a sync symbol (at worst 17 ms away) occurs. Since the noise has not changed, it is much smaller than the signal on any used tone. The newly estimated set of tone gains can be compared to the stored channel profile from the past. The receiver equalizer can be updated immediately in response to the newly calculated gain, and the resulting new bit distribution is immediately communicated to the transmitter via the ES command. The total estimation time may be about 20 ms before the new bit distribution is known.
[0062]
In the frequency domain on the nth tone, the output Y_nIs input X by the number 1_nAssociated with.
[0063]
[Expression 1]

[0064]
Where H_nAnd N_nIs channel gain and noise power. A set of pre-identified channel gains (based on pre-initialization, retraining, or fast swap) can be stored for each channel profile that the DMT system has pre-identified. These stored profiles are P if l = 1,._{l, n}It is represented by Since the DMT system clock is offset by a relative timing offset ε from the last time the profile occurred, it causes a linear phase shift from the old profile setting.
[0065]
[Expression 2]

[0066]
With a successful profile detection, a known X of the synchronization symbol is included in the prior profile._nH measured last for_nIt is identified whether there is a profile that best matches.
[0067]
First, the profile must be generated because the profile is not known. H_nIs a simple estimate.
[0068]
[Equation 3]

[0069]
Here, M is the number of used synchronization symbols. As M becomes larger, an accurate channel estimation value can be obtained, but the realization of the correction high-speed swap is delayed. Since the difference in estimates generally decreases in proportion to M, the measured SNR is too high or too low by the number 4 with 99% determinism (too high is a problem).
[0070]
[Expression 4]

[0071]
After 4 synchronization symbols, the calculated SNR error is less than 2 dB with 99% determinism. Any receiver gain / phase rotation (commonly referred to as “FEQ”) needs to be immediately adjusted by the ratio of Equation 5. And if necessary, the new channel can be stored as one of the profiles.
[0072]
[Equation 5]

[0073]
However, if a profile has been identified in advance, the synchronization signal can be used alone to detect the reoccurrence of this profile reliably with maximum likelihood profile detection, thus ( Possible vector profile P_lThe profile that minimizes the sum is selected.
[0074]
[Formula 6]

[0075]
The probability that a false profile is detected depends on the noise and the proximity between the profiles, and is particularly constrained by Equation 7.
[0076]
[Expression 7]

[0077]
What is important here is that if two profiles confused with each other are too close and the corresponding bit / gain table is functional, false detection does not correspond to a link malfunction. By maintaining the profile exactly with the receiver expert system, it is necessary to select the profile so that the corresponding correction bit distribution is very unlikely to cause false detection leading to DMT link failure.
[0078]
Further, if no significant channel gain (or output power) change is detected by the total power measurement, a noticeable noise change is assumed. In order to accurately estimate the noise difference, it takes several seconds in the DSL modem even when the training signal is used. However, as usual, if the margin before noise introduction is high, the subchannel can be loaded carefully. Gaussian noise samples typically fall within a range of ± 3σ (3 standard deviation) distribution points with a probability of 0.997. An increase in noise of 10 dB or more causes an unacceptably high error rate in a DSL modem that is originally operating at 1e-7 with a 6 dB margin. Regardless of whether there is a slicer error in the FFT output in the receiver, there is a greater difference in the distribution of error samples for large noise. The receiver can generally estimate its distribution (for Gaussian noise, this distribution estimates the difference). A considerable number of samples are needed to estimate the exact difference of the new noise very accurately (about 3200 samples are needed to obtain an accuracy of 1/10 dB, which is inefficient On the other hand, it is possible to train within one second even if a simple estimator is used.) On the other hand, it is possible to use a faster gloss indicator in response to a large change. Those tones where significant noise increases are detected are zeroed in the bit table via ES. As these tones are deactivated in response to ES requests, subsequent noise estimation can continue. If the loading result is more accurate, a more detailed bit distribution can be transmitted by subsequent ES commands. In addition, very fast adjustments via ES commands by first approximating the noise profile close to ISDN crosstalk, HDSL crosstalk, or other types of known crosstalk, and then refining with high performance receivers Is possible.
[0079]
More specifically, if the change in noise is large, transmitting no signal, that is, outputting 0 from the tone, is an effective method for determining the noise. In the following, several methods for estimating noise by output zeroization of tones will be described.
[0080]
In general, if there is no channel change and (assumed) synchronization symbols are known, an instantaneous noise sample for each tone can be determined. This is optimally represented by Equation 8.
[0081]
[Equation 8]

[0082]
The noise power is estimated by equation (9).
[0083]
[Equation 9]

[0084]
The receiver then notifies the transmitter via an ES command to zero these tones. Here, Equation 10 is larger than a specific threshold proportional to the original noise power.
[0085]
[Expression 10]

[0086]
In general, in order to speed up noise estimation, the receiver must select a small value as M. As a result, the noise power estimation becomes inaccurate, which may result in an inaccurate tone output zeroing profile. The accuracy of the tone output zeroing profile can be improved by assuming that the power spectral density of the noise difference is relatively smooth. There are many clustering methods that use such smoothness assumption.
[0087]
In one embodiment, the situation that defines a continuous tone set within a cluster is generally an end tone that is a zeroed or unused tone, and the tone that precedes and follows the cluster is the tone after use. It is. In this embodiment, only tones intended for use by the modem (ie, non-POTS tones) are targeted. In general, a zeroed tone is defined as a tone that transmits 0 bits due to a sufficiently large increase in noise, a used tone is defined as a tone that transmits non-zero bits after a new loading, and an unused tone is , Defined as a tone that was transmitting zero bits prior to loading, either due to low channel gain or high noise differences.
[0088]
In the following embodiment, a clustering method is shown. Here, the symbols 0, X, and U represent a zeroized tone, a tone after use, and an unused tone, respectively. For 20 tones, given an output zeroing profile for the following tones:
[0089]
[Table 1]

[0090]
The possible clusters are (2,20), (2,18), (2,9), (2,6), (4,20), (4,18), (4,9), (4, 6), (8, 20), (8, 18), (8, 9), (12, 20), (12, 18).
[0091]
The next step following cluster identification is to determine which cluster non-zeroing tones are zeroed. This can be done according to the following rule of thumb. First find the largest size cluster with at least a certain percentage of tones zeroed, then look for the next largest cluster with the same percentage of tones zeroed, until all clusters have been searched repeat. If the size of this cluster is the same as the number of tones after use, retain. Otherwise, zero the non-zeroed tones in the cluster, move bits to other tones after use if possible, ignore the already zeroed cluster and return to the first step. In the above-described embodiment, when the above-described ratio is 75%, the tones 3, 7, and 19 are zeroized.
[0092]
An alternative method of zeroing tones without having to send an ES command is to zero a subset of tones using a predetermined sequence that is known to both the transmitter and receiver. In this case, a command is required to signal the start of this tone zeroing method. For zeroed tones where the measured signal is only noise, a moving average estimate of the noise power can be obtained. For zeroed tones, if the signal average exceeds 40 symbols, the signal is within 2.0 dB of true noise difference with 99% reliability, or true with 99.9% reliability. The noise difference can be in the range of 3 dB.
[0093]
Once a set of noise estimates is obtained, it can be matched with the stored noise profile by the maximum likelihood method. Typical examples of crosstalk noise profiles include ISDN, HDSL, and T1.
[0094]
In using maximum likelihood estimation as the most appropriate profile selection method, it is necessary to find a noise difference profile that most closely matches the obtained noise sample. If all the profiles are similar, Equation 11 needs to be obtained.
[0095]
[Expression 11]

[0096]
Where Equation 12 is an L-dimensional complex vector containing noise measurements on the kth tone,_jIs the jth crosstalk profile including the difference. L indicates the number of the noise measurement performed on the kth tone. This equation (Equation 11) is simplified as Equation 13.
[0097]
[Expression 12]

[Formula 13]

[0098]
Where L is the number of symbols used to estimate the noise difference and σ² _{j, i}Is the difference of the i-th tone in the j-th profile, which is the number 14, the number 15,_{k, i}Is a measure of the noise at time k of the i th profile, and D_iIs the empirical noise difference on the i th tone. The above equation (Equation 13) makes it possible to identify the noise profile closest to the noise measurement.
[0099]
[Expression 14]

[Expression 15]

[0100]
The following three examples compare the realization speed of high-speed swapping and bit swapping.
[0101]
Example 1: Swapping one bit
If one bit is moved by conventional bit swapping, an AOC bandwidth of 60 bytes is required. This can logically traverse a bi-directional AOC channel in approximately 30 ms (60 bytes / (2000 bytes / second)). However, G. Wright and G. In the standard of dmt, swap conversion of 1 bit or more cannot be performed within 800 ms (due to implementation delay). 800 ms does not apply to fast swapping. Due to reverse interoperability, it is difficult to eliminate this 800 ms problem from conventional bit swapping (however, such a problem does not exist for modems using high speed swapping). Conversely, fast swapping requires an 8 byte command that uses only a unidirectional AOC channel, so approximately 21 ms (8 bytes / (2000 bytes / second)) + 17 ms (worst case timeout) time. It takes.
[0102]
[Table 2]

[0103]
Example 2: Moving 40 bits on 10 tones
This type of swap is typically a gradual noise change on the line, for example when the AM radio receiver power level up at night. The conventional bit swapping logically requires at least 60 bytes × (40 bits) = 2400 bytes, and takes 1.2 seconds. The actual time taken is 40 (0.8 s) = 32 seconds (assuming a minimum of 800 ms is required for each 1-bit swap). In contrast, high-speed swapping requires 2 bytes × (10 tones) +4 bytes = 24 bytes, and takes 29 ms (including the worst-case timeout of 17 ms). Therefore, the speed increases as the number of bits increases.
[0104]
[Table 3]

[0105]
Example 3: 400-bit profile change on 100 tones
This type of swap is typically a drastic channel change on the line, such as an off-hook impedance change. The bit swapping protocol requires 12 seconds logically and 320 seconds in practice. In this case, it is likely that retains are obviously faster and need to be used, as errors can occur while correcting the problem by bit swapping. It takes 119 ms for fast swapping to move all 400 bits with a 204 byte AOC command. Bit-swapping is significantly faster and is significantly faster than retain because there is no service intervention (however, if the channel characteristics change significantly, it may cause some bit errors in the fast-swapping interval) ).
[0106]
[Table 4]

[0107]
More specifically, to realize ES, (2n + 4) / R_AOC+ T_time-outSeconds are required, where n = number of tones, R_AOC= AOC channel speed in bytes / second, T_time-out= Timeout time. The timeout period is 17 ms in the worst case and may be much faster. The time-out period is the time when the receiver does not detect any change in the bit table of the received channel output after the AOC byte was last transmitted. Is the time required to assume that it was not possible or impossible to realize. For simplicity, the timeout is chosen to be the next superframe boundary (but logically smaller). In the conventional swapping, when the number of moved bits is b, the required time is only 0.8 b seconds. FIG. 7 is a graph in which the improvement degree of b = kn is plotted with respect to the number of tones for each case of k = 1, 2, 5, and 10. (This particular plot actually uses a 2ms timeout, which is still within the scope of interest) As shown, the improvement in ES protocol is very large in all cases. The protocol itself is therefore capable of performing high speed swapping.
[0108]
As can be seen from the above, the present invention has many advantages over the prior art. For example, bit swapping is accelerated by high-speed swapping. In some cases, the speed of high-speed swapping can reach 10,000 times that of bit-swapping. Furthermore, the execution time of fast swapping is almost instantaneous and is sufficient to maintain connectivity between a number of significant changes, especially when a large cross talker is suddenly introduced. Also, using fast swapping commands does not require passing additional information during initialization. That is, there is no confirmation notification packet, and the change request command need only be transmitted once. Furthermore, fast swapping can reduce the need and frequency of retraining by increasing the reliability of the connection in situations where faster changes are required.
[0109]
Details related to bit swapping are also described in the following references: Salvekar et al. “CHANNEL GAIN CHANGE DETECTION AND CHANNEL PROFILE SELECTION IN A MULTCARRIER SYSTEM”, Globecom, December 1999, John M. Cioffi “VDSL ALLIANCE: VDSL TRANSMISSION SPECIFICATION (VDSL transmission specification) (00-116) "T1E1.4: VDSL, February 2000, and Cioffi et al." G.LITE.BIS: ANALYSIS OF EXPRESS SWAPPING SPEEDS AND RELIABILITY Speed and Reliability Analysis) ”April 1999, which are incorporated herein by reference.
[0110]
Next, an alternative method for increasing the speed of swap will be described. This new method includes a receiver and a transmitter that both monitor the channel and store various channel change profiles. If a change is detected during processing, the receiver and transmitter switch to the same stored profile (one of 16 profiles) and continue operation. Therefore, no high-speed swap command or other commands are transmitted from the receiver, so that a faster swap is performed. More specifically, the receiver and transmitter are configured to include a line monitor for monitoring channel performance and controlling signal adaptation. For example, the receiver and transmitter may include a line monitor as described above. The receiver and transmitter are also configured to store a plurality of predetermined profiles corresponding to specific line changes. Thus, if a change in the channel is detected by the monitor, the receiver and transmitter instruct the associated device to select a predetermined profile that best matches the desired change. In this way, both the receiver and the transmitter can operate with a predetermined profile. For example, the predetermined profile may include a bit and gain distribution for each tone of the multitone signal.
[0111]
While several embodiments have been described in detail above, it should be understood that the present invention can be implemented in many other specific forms without departing from the spirit and scope of the present invention. The invention has been described with particular reference to fast swapping for discrete multitone systems. However, it should be understood that the present invention can be used in various other systems. For example, multi-channel or multi-carrier systems that require dynamic adaptation of sub-channels (sub-carrier bits and gain). Furthermore, the monitoring of the receiver can be realized in various ways, and it is considered that significant changes can be made without departing from the scope of the present invention. The embodiments described above are therefore considered to be illustrative and not restrictive, and the invention is not limited to the details listed here, but can be varied within the scope of the appended claims.
[Brief description of the drawings]
FIG. 1 is a diagram illustrating a format of a T1.413 bit swapping request.
FIG. 2 shows a typical bit and gain table.
FIG. 3 is a diagram illustrating a format of a high-speed swapping request according to an embodiment of the present invention.
FIG. 4 illustrates a modem architecture according to one embodiment of the present invention.
FIG. 5 illustrates a receiver in accordance with one embodiment of the present invention.
FIG. 6a shows a typical output constellation for a 4-bit allocation tone.
FIG. 6b illustrates a typical output constellation for a 6-bit allocation tone.
FIG. 7 is a graph plotting improvement factors for fast swapping versus the number of tones affected.
[Explanation of symbols]
12 ... Message header
14 ... Message
16 ... Message field
18 ... Command
20 ... Tone index
50 ... Bit and gain table
52 ... Tone Number Field
54 ... Bit allocation field
56 ... Gain or power level field
100 ... High-speed swapping command
102: Message header field
104 ... High-speed swapping control field
106 ... tone message
108 ... Error field
110: Tone index
112 ... Desired value indicator
200 ... modem
202 ... Transmitter
204 ... Receiver
206: Encoder
208: Modulator
210 ... Window operation filter
211 ... Controller
212 ... Analog interface
214 ... Analog interface
216 Time domain equalizer
218 ... demodulator
220: Decoder
221 ... Controller
222: Remote device
223 ... Transmission media
224 ... Bit and gain table
226 ... ADSL overhead channel
227 ... Old bit / gain table
228 ... New bit / gain table
300 ... DMT receiver
302... First decoder
304: Second decoder
305 ... Syndrome generator
306 ... Syndrome generator
308 ... Switch
400 ... 4-bit placement
500 ... 6-bit placement

Claims (30)

A method for adaptively changing communication signal parameters in a transmission system using multi-carrier modulation,
The change is made under the condition that the communication signal does not include a response defined as indicating that a change request has been received or realized;
A process for transmitting a change request from the receiver to the transmitter, the change request identifies individually specific subcarrier to be changed, the individual desired values of the identified parameters associated with each sub-carrier has been is configured to identify the respective desired value of the parameter is different from the current value is the current value of the parameter, the steps,
Monitoring the received communication signal and determining whether the requested change has been realized, wherein determining whether the requested change has been realized includes determining whether the requested change has been realized. A method comprising: a step based at least in part on an analysis of a portion intended to be changed. The method of claim 1, comprising:
The change request is configured to identify a plurality of specific subcarriers to be changed and to identify a desired value of a parameter associated with that subcarrier for each identified subcarrier. A method according to claim 1 or claim 2 , wherein
The monitoring step decodes the received communication signal using the current value of the parameter and the desired value of the parameter, and the requested change is based at least in part on the decoding. Determining whether it has been realized. The method of claim 3 , further comprising:
Whether or not the change request has been realized, the step of generating a first forward error correction syndrome based on the decoding of the current value and a second forward error correction syndrome based on the decoding of the desired value Determining the step comprising: making a determination based on an analysis of the syndrome. A method according to any one of claims 1 to 4 , comprising:
The method of monitoring includes monitoring the particular subcarrier intended to be changed by the change request and determining whether the requested change has been realized. A method according to any one of claims 1 to 5 , comprising
The determination of whether the requested change has been realized is at least partly in analyzing an error detected based on decoding the communication signal using the current value and the desired value of the parameter. Based on the method. A method according to any one of claims 1 to 6 , comprising
The step of monitoring includes analyzing a first placement of the subcarrier at a first time point by a second placement of the subcarrier at a second time point;
The method, wherein the change request is known to be realized when the second configuration is different from the first configuration. A method according to any one of claims 1 to 7 , comprising
The monitoring step includes monitoring one of the subcarriers of the communication signal, and the subcarrier has been changed to be equal to 0 based on the realization of the change request;
A method wherein a lack of transmitted energy on the subcarrier indicates that the requested change has been accomplished by the transmitter. A method for facilitating bi-directional communication between a pair of modems using an adaptive multi-carrier transmission signal, comprising:
Transmitting a change request command from a first modem to a second modem, the change request command identifying a specific subcarrier to be changed, to a specific subcarrier in the multicarrier transmission signal; Configured to communicate a desired change in an associated parameter value, wherein the desired change in the parameter value is a change to a value different from the current parameter value;
Monitoring a communication signal received at the first modem to determine whether the change requested by the change request command has been realized, whether the requested change has been realized; A determination is made of a portion of the received communication signal intended to be changed, provided that the communication signal does not include a response defined as indicating that the change request has been received or realized. A process performed based at least in part on the analysis; and
With
The change request command is:
A header configured to identify the command as a change request command;
A control field configured to indicate the number of tones changed by the change request command;
At least one subcarrier identifier, each configured to identify a particular subcarrier to be changed by the change request command;
At least one desired parameter value indicator, each configured to identify a desired value of a parameter associated with the subcarrier identified by the associated subcarrier identifier;
A method comprising an error field. The method of claim 9 , comprising:
The method wherein the desired parameter value indicator identifies the number of bits of information transmitted by the subcarrier identified by the associated subcarrier identifier. The method of claim 10 , comprising:
The method wherein the desired value of the indicator is a total number of bits changed for a particular subcarrier. The method of claim 9 , comprising:
The method wherein the desired parameter value indicator identifies a gain used by the subcarrier identified by the associated subcarrier identifier. The method of claim 12 , comprising:
The method wherein the desired value of the indicator is incremented by about 0.5 dB in the range of about -4.0 dB to about +3.5 dB. A method according to any one of claims 9 to 13 , comprising
The method, wherein the error field includes a cyclic redundancy check code. 15. A method according to any one of claims 9 to 14 , comprising
The method, wherein the header includes a superframe number indicating when the change request is made. 15. A method according to any one of claims 9 to 14 , comprising
The method, wherein the control field includes a superframe number indicating when the change request is made. A method for adaptively changing communication signal parameters in a transmission system using multi-carrier modulation,
Receiving a change request from a receiver at the transmitter, the change request identifying a set of specific subcarriers to be changed, and at least one parameter associated with each of the identified subcarriers; Are configured to identify a corresponding set of desired values, wherein the desired value is different from a current value, which is a current value of the parameter, and each desired value of the set of desired values is Applying to change each of the identified subcarriers , wherein the request for change includes a subcarrier count configured to indicate a number of subcarriers to be changed by the change request; and
Changing the identified subcarrier to maintain the desired value of the parameter associated with the identified subcarrier by implementing the request for modification;
Transmitting a multi-carrier signal that includes the realized change request and does not include a response command defined to indicate that the change request has been received or realized . The method of claim 17 , comprising:
The change request includes a header configured to identify the change request. 19. A method according to claim 17 or claim 18 , wherein
The request for change includes at least one subcarrier identifier, each configured to identify a particular subcarrier to be changed by the change request. A method according to any one of claims 17 to 19 , comprising
The request for change includes at least one indicator, each configured to identify a desired value of a parameter associated with a subcarrier identified by an associated subcarrier identifier. A method according to any one of claims 17 to 20 , comprising
The method, wherein the change request includes an error field. A method according to any one of claims 17 to 21 , comprising
The method, wherein the change request includes a superframe number indicating when the change request is made. A method for performing high-speed bit swapping in a discontinuous multitone transmission system, comprising:
Transmitting a change request from a receiver, wherein the change request is configured to identify a particular tone to be changed and to identify a desired value of a parameter associated with the identified tone; The value is different from the current value, which is the current value of the parameter, and the process;
Receiving the change request at a transmitter, wherein the change request is configured to identify a particular tone to be changed and to identify a desired value of a parameter associated with the identified tone; The desired value is different from the current value of the parameter; and
Modifying the identified tone to maintain the desired value of the parameter associated with the identified tone by implementing the request for modification at the transmitter;
Transmitting a multi-tone signal including the realized change request from the transmitter;
Receiving the multitone signal at the receiver;
Interpreting the received multi-tone signal based at least on the desired value of the parameter and determining whether the requested change has been realized. 24. The method of claim 23 , comprising:
The method, wherein the change request is configured to affect a substantial number of tones. 25. A method according to claim 23 or claim 24, comprising:
The method, wherein the multitone signal does not include a response command defined to indicate that the change request has been received or realized . A discontinuous multitone modem communicating with a transmitter,
A receiver configured to transmit a change request to the transmitter, the change request identifying a particular tone to be changed and identifying a desired value of a parameter associated with the identified specific tone The receiver is configured to receive a multi-tone signal from the transmitter and to determine whether the change request has been realized by the transmitter, and whether the change request has been realized. The determination is based at least in part on an analysis of a portion of the multitone signal that is intended to be modified,
The determination of whether or not the change request has been realized does not include a response defined as indicating that the change request from the transmitter has been received or realized that the change request has been realized ,
The desired value of the parameter is different from a current value which is a current value of the parameter;
modem. 27. A modem according to claim 26 , wherein
The receiver is configured to analyze a plurality of received first tones and a plurality of received second tones, wherein the first tone is a parameter associated with the first tone. The second tone includes a parameter changed based on the change request, the change request is implemented at the transmitter, and the receiver sets the parameter of the first tone to the Configured to determine whether the first tone is different from the second tone by comparing with the modified parameter of a second tone;
If the parameters of the first and second tones are substantially the same, the change is not realized, and if the parameters of the first and second tones are not substantially the same, the change Is realized, a modem. 27. A modem according to claim 26 , wherein
The receiver comprises a first decoder and a second decoder, the first decoder comprising a plurality of first parameter values associated with a first tone set, the second decoder comprising: A plurality of second parameter values associated with a second tone set, wherein the first decoder comprises a first syndrome generator, the second decoder comprises a second syndrome generator, and these syndromes The generator is configured to calculate a first syndrome associated with the first syndrome generator and a second syndrome associated with the second syndrome generator, wherein the first syndrome is received A plurality of tone parameter values assigned to the first tone set associated with the first tone set; Configured to compare with a parameter value, wherein the second syndrome is configured to compare a parameter value of the received plurality of tones with the second parameter value associated with the second tone set. Configured modem. A modem according to any one of claims 26 to 28 , comprising:
The receiver is configured to monitor one of the tones of the communication signal and analyze a first configuration of the tone at a first time point by a second configuration of the tone at a second time point. ,modem. 30. A modem according to any one of claims 26 to 29 , comprising:
The modem , wherein the receiver includes a monitor that monitors one of the subcarriers of the communication signal and the subcarrier is changed to be equal to zero based on the realization of the change request.

Images