200539601 玖、發明說明： 【發明所屬之技術領域】 本發明是關於解決正交分頻多工系統(Orthogonal Frequency Division Multiplexing，OFDM)發射端之高峰 均值（Peak>to-Average Power Ratio，PAPR)的技術， 特別是一種降低OFDM信號PAPR值的方法與裝置。200539601 发明 Description of the invention: [Technical field to which the invention belongs] The present invention relates to solving the peak-to-average power ratio (PAPR) of the transmitting end of the Orthogonal Frequency Division Multiplexing (OFDM) system. Technology, particularly a method and device for reducing the PAPR value of an OFDM signal.
【先前技術】[Prior art]
常使用的離散多載波轉換(Discrete Multitone，DMT) 及OFDM之多載波通訊系統在高速通訊應用中吸引了廣 大的注意，這些高速通訊應用諸如：數位用戶迴路(Digital Subscriber Line，DSL)、數位地面廣播(digital terrestrial broadcasting)、無線區域網路（Wireless Local Area Network，WLAN)、無線都會網路(Wireless Metropolitan Area Network，WMAN)、專用短距離通信系統(Dedicated Short Range Communication，DSRC)以及電源線通訊(power line communication)等等。它們也可望成為下一世代行動通 訊的主流。多載波通訊系統的優點來自於將一高速資料流 (datastream)分隔成為多重平行的資料流，而這些多重平行 的資料流是藉由個別副載波(subcarrier)來傳送。因此，每 一部分資料流是以低速傳輸，具有較強的抗多路徑通道 (multipath channel)效應及窄頻干擾(narrowband interference) 的能力。 7 200539601 第一圖為傳統多載波通訊系統有關OFDM發射機的 系統方塊圖。在OFDM發射機中，輸入資料；^]，灸=〇,丨， AM，係傳送於一 0FDM符元(symbol)週期之内，經串/並 (Serial/Parallel，S/P)轉換後，藉由#點反向快速傅立葉轉 換(Appoint Inverse Fast Fourier Transform ’ iV-IFFT)，再經並 /串(Parallel/Seria卜P/S)轉換成為以下的離散時間序列： 其中 WN^ej2n,N (2) 為轉動因子(twiddle factor)。由⑴式獲得的離散時間序列 对《]接著經過循環前置插入(cyclic prefix insertion)後，進行 數位/類比轉換獲得一類比信號x(〇。產生的類比信號χ(ί) 再被傳送到RF前端作進一步處理，包含IQ調變(IQ modulation)、升頻轉換(up conversion)和功率放大(p0Wer amplification)等。類比信號χ(ί)的PAPR值較高於對應的離 散時間信號x[w]之PAPR值數個dB，且大約近似;φα]之 PAPR值。此χ〇ζ/Λ]表示由χ[η]之Λ倍超取樣(oversampiing) 所取得的序列。因此，由x[«/及]可得x(〇之近似PAPR值為 max \x[n/R]\2 PAPR = τ— (3) E{\x[n/R)\2} 其中五{·}表示期望值的運算。一般在Λ > 4時，此近 似值是相當準確。然而，此類習知多載波通訊系統主要缺 200539601 點之為調變^號的高PAPR值。當高PAPR值的調變信 號經過好前端時，此信號會因為-般RF功率放大器的非 線性特性而受敎真，此祕性特性不僅會造成頻帶中 (^^㈣仏號失真而導致誤碼率出红^叮沉尺拙^^^升 高，而且會引起該頻帶外能量散逸(out-of-band radiation)(或 頻譜再生)而導致相鄰的頻道干擾及違反政府頻譜規定。此 問題常見的解決方法是簡單地糊具有較大線性範圍的功 率放大器，但會導致功率效能降低、較高的功率消·耗量和 較高的製造成本。 解決此問題已經有許多習知的方法。這些方法包括區 塊編碼(block coding)、截掉(dipping)、部分傳輸序列(Partial Transmit Sequences ’ PTS)、選擇性對應(Selective Mapping， SLM)、載波保留（Tone Reservation，TR)、載波注入(Tone Injection ’ Ή)及脈衝重整(pulse superposition)等方法。在這 些方法中，PTS方法在實施複雜度與PAPR降低性能上最 具有吸引力，Ericsson公司在美國專利6，125，103中揭露了 一種使用PTS方法解決OFDM發射端信號之高PAPR值的 問題，其方塊圖如第二圖所示，說明如下。 首先將長度#的輸入資料別幻在頻域(frequency domain)上劃分為Af個分離的子區塊(或組），表示為不⑷、 為闲、…、义4幻，灸=〇, 1，..·，ΛΜ。該劃分方式可以是交 錯(interleaved)、鄰接(adjacent)和不規則(irregular)等方式， 200539601 如第三圖所示(以你=4為例）。此Μ個分離的子區塊接著 被相位旋轉且相加在一起以形成下列信號： 戈W = [fc]，fc = 〇, 1，…，W -1 (4) m=l 其中心是有關於第W個子區塊(we {1，2,…，州)的相位旋 轉參數(註，丨心丨=1)。 (4)式經MIFFT產生： 〜 μ Σ ㈨]，η = 0,1，…，W — 1 (5) m=l 其中表示對尤阅作at_ifft運算之結果。在PAPR降 低演算法中，PTS方法的目標是相位最佳化，亦即尋找最 理想的組合序列{匕，心，…，，使得對應的發射信號的 PAPR值為最小。在實際應用上，心的相位通常會限制為 最多只有四種可能的數值{+1，-1，勺·，W，如此相位旋轉動作 便不需要任何乘法運算。 從第二圖的PTS實現方法中可窺知一個見點的OFDM 符元需要做Μ次iV-IFFT運算，亦即需要M)1〇g2w個複 數乘法運算’因此Kang，Kim與J〇〇以及Samsung公司都 提出了降低PTS運算量的實現方法。Kang，與J〇〇論 文中（A novel subblock partition scheme for partial trammit sequence OFDM，IEEE Trans· Broadcasting，vol· 45, no, 3, 200539601 PP.紐·说細·聊）提到的方法可用第四圖來表示(以 為例），其原理為利用PTS的交錯劃分區塊特性，在 頻域上每一個子區塊共有#點，但只有£點有值 均’其餘為0，因此將此TV點子區塊信號尤闲做#_IFFT 運算，相當於對一個Z點子區塊(不中不為〇值的資料） k號經過Z-IFFT ’並在時域(time domain)上重覆从次形成 #點的信號，再將此#點信號乘以#點的複數係數 細-V，m = 〇1…，M-1， 此方法的運算量為+ ·個乘法數目，所需要的 記憶體為胃個單元。The commonly used Discrete Multitone (DMT) and OFDM multi-carrier communication systems have attracted wide attention in high-speed communication applications such as: Digital Subscriber Line (DSL), digital terrestrial Digital terrestrial broadcasting, Wireless Local Area Network (WLAN), Wireless Metropolitan Area Network (WMAN), Dedicated Short Range Communication (DSRC), and power line communication (power line communication) and so on. They are also expected to become the mainstream of mobile communications for the next generation. The advantage of a multi-carrier communication system comes from separating a high-speed data stream into multiple parallel data streams, and these multiple parallel data streams are transmitted by individual subcarriers. Therefore, each part of the data stream is transmitted at a low speed, and has a strong ability to resist multipath channel effects and narrowband interference. 7 200539601 The first figure is a system block diagram of an OFDM transmitter in a conventional multi-carrier communication system. In the OFDM transmitter, input data; ^], moxibustion = 0, 丨, AM, are transmitted within a 0FDM symbol period, and after serial / parallel (S / P) conversion, Through #point Inverse Fast Fourier Transform (iV-IFFT), and then parallel / serial (Parallel / Seria, P / S) conversion into the following discrete time series: where WN ^ ej2n, N ( 2) is the twiddle factor. The discrete time sequence pair obtained by 《] is then subjected to cyclic prefix insertion, and then subjected to digital / analog conversion to obtain an analog signal x (〇. The generated analog signal x (ί) is then transmitted to RF The front end performs further processing, including IQ modulation, up conversion, and p0Wer amplification. The analog signal χ (ί) has a higher PAPR value than the corresponding discrete-time signal x [w ] PAPR value of several dB, and approximately approximate; PAPR value of φα]. This χ〇ζ / Λ] represents the sequence obtained by Λ oversampiing of χ [η]. Therefore, from x [« / And] It can be obtained that x (〇 has an approximate PAPR value of max \ x [n / R] \ 2 PAPR = τ— (3) E {\ x [n / R) \ 2} where five {·} represent the expected value of Operation. Generally, this approximation is quite accurate when Λ > 4. However, this conventional multi-carrier communication system mainly lacks the high PAPR value of the modulation ^ number 200539601. When the modulation signal of the high PAPR value passes well, At the front end, this signal will be affected by the non-linear characteristics of the general RF power amplifier. This mysterious characteristic will not only cause (^^ ㈣ 仏The distortion of the signal causes the bit error rate to be red ^ ding Shen ^ ^ ^ ^ ^ increased, and will cause out-of-band radiation (or spectrum regeneration) outside the band, resulting in adjacent channel interference and violations Government spectrum regulations. A common solution to this problem is to simply paste a power amplifier with a large linear range, but it will lead to reduced power efficiency, higher power consumption and consumption, and higher manufacturing costs. There are already solutions to this problem Many conventional methods. These methods include block coding, dipping, partial transmission sequence (PTS), selective mapping (SLM), carrier reservation (Tone Reservation, TR), carrier injection (Tone Injection 'Ή), and pulse superposition. Among these methods, the PTS method is the most attractive in terms of implementation complexity and PAPR reduction performance. Ericsson Corporation in the United States Patent 6, 125, 103 discloses a method using PTS to solve the problem of high PAPR value of the OFDM transmitting end signal. The block diagram is shown in the second figure, which is explained as follows. The input data of # is divided into Af separated sub-blocks (or groups) in the frequency domain, which is expressed as ⑷, idle, ..., meaning 4 magic, moxibustion = 〇, 1, ... ·, ΛΜ. The division method can be interleaved, adjacent, and irregular. 200539601 is shown in the third figure (take you = 4 as an example). The M separated sub-blocks are then phase rotated and added together to form the following signal: Go W = [fc], fc = 〇, 1, ..., W -1 (4) m = 1 and its center is Regarding the phase rotation parameters of the W-th sub-block (we {1,2, ..., state) (note, 丨 heart 丨 = 1). (4) The formula is generated by MIFFT: ~ μ Σ ㈨], η = 0,1, ..., W — 1 (5) m = l, which represents the result of the at_ifft operation for the special reading. In the PAPR reduction algorithm, the goal of the PTS method is to optimize the phase, that is, to find the optimal combination sequence {dagger, heart, ..., so that the corresponding PAPR value of the transmitted signal is minimized. In practical applications, the phase of the heart is usually limited to a maximum of four possible values {+1, -1, spoon ·, W, so that the phase rotation action does not require any multiplication. From the PTS implementation method in the second figure, it can be seen that an OFDM symbol of a viewpoint needs to perform M iV-IFFT operations, that is, M) 10 g2w complex multiplication operations. Therefore, Kang, Kim and J〇〇 and Samsung companies have proposed implementation methods to reduce the amount of PTS calculations. The method mentioned in the paper by Kang and JOO (A novel subblock partition scheme for partial trammit sequence OFDM, IEEE Trans · Broadcasting, vol · 45, no, 3, 200539601 PP. New · Detail · Talk) can be used in the fourth Figure to show (for example), the principle is to use the PTS's interleaved block characteristics, each sub-block in the frequency domain has a total of # points, but only £ points have a value of 'the rest are 0, so this TV idea The block signal is particularly idle to do #_IFFT operation, which is equivalent to a Z-point sub-block (data not equal to 0). The k number passes through Z-IFFT 'and is repeated from time to time to form # Point signal, then multiply the #point signal by the # complex number of points -V, m = 〇1 ..., M-1, the calculation amount of this method is + · number of multiplications, the required memory is stomach Units.
Samsung公司在美國專利申請案US 2003/0067866也 提出類似的觀念，如第五圖所示。和前述方法不同的是每 一個i：點子區塊經過Z_IFFT後不再重覆，直接在時域乘上 Z點的複數係數使形成的時域子區塊資料彼此正交，以方 便接收端分離出各個子區塊資料。由於各個時域子區塊只 有Z點，其PAPR值較低，所以經過相位旋轉後加總所得 的發射信號的PAPR值也較低。雖然此方法所需的乘法數 目為M.〇〇k)g2L + JV，所需要的記憶體為#個單元，但是 此方法將原本長度#的OFDM發射信號縮短為長度jL的 發射信號，意謂著此系統對抗多路徑通道效應的能力也會 隨之降低許多。而且在許多情況下根本無法設計出I點的 複數係數乘法器來使得發射端形成的時域子區塊資料彼此 200539601 正父如此將造成接收端不易還原出原本傳送的資料。 【發明内容】 為克服上述傳統OFDM發射端降低PAPR值之 PTS實現方法的缺點，本發明主要目的為提供一種降低 OFDM信號PAPR值的方法與裝置。本方法利用pTs 交錯劃分區塊的特性，在時域上直接將長度iV的一離散時 間序列X间劃分為複數個不互相重疊的子區塊後，·再經轉 換組合及一相位最佳化處理而合成一完整的#點信號 对β，其中#為一正交分頻多工信號的長度且” =〇, 1，…，ΛΜ 〇 本發明只需用到一個7V4FFT，可大幅降低運算量，所 需複數乘法數目為(M2)log2#，所需要的記憶體為#個單 元，更重要的是本發明仍保有OFDM原本具有的抗多路徑 通道效應之能力。 本方法更包含下列步驟:首先將長度1的整數) 的一離散時間序列对《]劃分為Μ個不互相重疊(disjoint)的 子區塊(或組），每一子區塊的長度為AT/#，M為2的次幂 (power)，且W/M為大於1的整數。接著，再利用一個組合 器(combiner)將此Μ個不互相重疊的子區塊合成另外μ個 長度的序列am，其中^=1，2,…，λ/且„ = 〇, 1，…， 。最後，運用對稱性質，將此μ個序列％卜]經相位 12 200539601 旋轉且相加在一起，形成一完整的TV點傳送信號3f[n]。 本方法中’第一和第二較佳實施例分別以Μ =2和Μ 的情況來說明上述之時域實現方法的步驟。 依此，本發明之降低0FDM信號pAPR值的裝置 主要包含一#點反向快速傅立葉轉換、一解多工器 (de-multiplexer)、一組合器、一組記憶體、和一個加法 器。#點反向快速傅立葉轉換將輸入的頻域信號；轉換 參 成時域信號χ[«]，解多工器將X间依鄰接方式劃分為从個 不互相重疊且長度相等的子區塊。組合器將此从個不互相 重疊的子區塊合成另外Μ個長度皆為；V/M的離散時間序 列《ηΙΧΙ。運用對稱性質，將此Μ個序列八[„]經相位旋轉 後，經由加法器相加而形成一完整的#點信號。在此 Μ個不互相重叠且長度相等的子區塊和从個序列乃[”]則 先後儲存在這一組記憶體裡。 以Μ=4的情況，本發明與其他三種PTS實現方法的 運算置與5己憶艘需求相比較。結果顯示，相較於原來及 Kang、Kim與Joo之PTS實現方法，本發明與Samsung之 PTS實現方法所需的乘法數相同且為最少，並且其記憒體 需求單元相同且為最少。而本發明不需要將0FDM的信號 長度縮短，因此仍保有OFDM系統原本具有的特性和優 。 13 200539601 效配口下删式、貫施例之詳細說明及_請專利範 圍，將上述及本發明之其他目的與優點詳述於後。 【實施方式】 第六圖為本發明之降低0FDM信號PAPR值的方 法的示意圖。首先’將長度#的頻域信號_做#_IFFT 6〇1運算得到長度TV的時域信號制，再將制依鄰接方 式劃分為Μ個不互相重疊的子區塊，如步驟6〇3所示，其 中母一子區塊的長度為AWl/，Μ為2的次幂，且八/从為大 於1的整數。接著，此Μ個不互相重疊的子區塊經組合器 605形成Μ個長度為#的離散時間序列％[„]，乃问，..， 。最後，此Μ個離散時間序列乃问經步驟6〇7之相 位旋轉且相加在一起，形成一完整的#點傳送信號对⑴： Φ] = bxyx [n] + b2y2 [n]^-^bMyM [n] ⑹ 由於序列外[«]具備對稱性質，因此組合器605只需形成長 度剔^的序列M〇]，h[i]，···aKAWH]}即可。 依此’第七圖說明了本發明之降低OFDM信號PAPR 值的裝置的一個結構示意圖。此降低OFDM信號 PAPR值的裝置主要包含一尽IFFT 601、一解多工器 70卜一組合器703、一組記憶體705、和一個加法器707。 MIFFT 601將輸入的頻域信號义[幻轉換成時域信號对^， 200539601 解多工器701將jcM依鄰接方式劃分為Μ個不互相重疊且 長度相等的子區塊。組合器703將此Μ個不互相重疊的子 區塊組合而成Μ個長度#/Μ的序列％[«]，其中^=1，2,…， Μ且《 = 〇, 1，···，(Λ「/Μ)小此Μ個序列w[w]經相位旋轉後， 經由加法器707相加而形成一段長度AMI/的傳送信號{?[〇]， 对1]，···，对(#/Μ)-1]}。藉由更換不同的相位旋轉參數&， 可再由此Μ個的序列办]求得下一段傳送信號{?[#/#]， 对(#/M) + i]，…，对(π/Μ)-ΐ]}。依此方式，最後寸求得全 部的傳送訊號？[«]。其中，Μ個不互相重疊的子區塊和Μ _ 個序列八[«]則先後儲存在記憶體705裡。 以Μ==2為例來作說明，如第八圖所示。序列尤问，w== 〇, 1，···，ΛΜ經解多工器701分成長度#/2的兩個子區塊 {4〇】，41]，…，X[(7V72) - 1]}和{χ[ΛΓ/2]，对(7V72)+1]，…，X[7V- ’儲存於記憶體705。組合器803將這兩個長度#/2 的子區塊組合形成下列兩個長度#的離散時間序列： 馨 yi [^] = x[n] + x[((n - Ν/2))ν] [^] = x[n] - x[((n - N/2))n ] ' (7) 其中(（·）V表示循環位移(circular shift ) #點，w = 〇, i，…， (#/2)·ι。組合器8〇3的結構相當簡單，以兩個加法器，⑼知 和803b，來實現第(7)式，其結構與2_IFFT兩相同。可以 证月⑺式中所得的離散時間序列乃问與乃问是等同於第 —圏中原本PTS方法所產生的離散時間序列Χι间與。 由(7)式可推得 15 200539601 [((« + f»N 1= [((w -f))N]^ x[((n ~ N/2))n ] + x[n) = ^ [n] 少2 [((” + f ))n ]= h [((” - f ))」=x[((” —寧))"卜刺=—a 间(8) 換言之， 乃[I]=少1 [〇]，少1 [I + U =少1 [1]，…，％ [iV -1] = % [I 一 1] h [号]=[〇]，h [f+1]= -少2 [l]，· · ·，少2 [at 一 η = 一h [夸—η (9) 因此組合器803只需形成長度#/2的序列(^[ο]，％[η， 乃[(#/2)-⑴和卜吼乃⑴，···，乃，/]) - 1]}即可。由於 {χ[0]，χ[1]，"·，χ[#-1]}不再需要，因此記憶體7〇5可釋 放出來供{^肌乃⑴，- 1]}和[y2[〇]，j；2[i]， 少2[(#/2) - 1]}使用。換句話說，所需要的記憶體為#個單 元。最後由⑹式及⑼式可得 x[n^f\^bxyx[ri\-b2y2[n] (1 〇) 其中《 = 0, 1，"·，(Μ2)-卜將(10)式進一步表示成 咖+手]Ρ = 〇，ΐ，《 = 0，1，···，|一 1 (11) 其中相位旋轉參數及\如第九圖所示。由第八圖可看 出’當&為+1，-1，+j或-j時，本發明總共所需的乘法運算 量來自於#_IFFT，即需要(M2)log2#個複數乘法運算，所 需記憶體為#個單元。當M=4的情況時，其實施示意圖 如第十圖所示。序列:φ?]，w = 0, 1，···,ΑΜ經解多工器701 分成長度Λ『/4的四個子區塊，{χ[0]，···，χ[(Λ//4)_ 1]}、{χ[# 200539601 /4]”.”x[(iW2)- 1] }、{x[TW2],..”x[(37W4)· 1] }、和{x[3iV /4]，···，χ[Λ^ 1] }。組合器1003將這四個長度τν/4的子區塊 組合形成下列四個長度#的離散時間序列： y} [η] = x[n] + - N/2))n ] + x[((n ~ N/4))n ] + x[((n - 3N/4))n ] y3 [n] = x[n] + x[((n - N/2))n ] - x[((n ~ N/4))n ] - x[((n - 37^/4))^ ] y2 [n] = x[n] - x[((w - N/2))n ] + jx[((n - N/4))n ] - jx[{{n - 37V/4))^ ] yA [n] = x[n] - x[((n - N/2))n ] - jx[((n - N/4))n ] + jx[((n - 3Λ^/4))^ ] (12) 組合器1003以八個加法器及一個虛數j的乘法器來實 現第(I2)式’其結構與4_IFFT兩相同。可以證明(u)式中 Φ 所得的離散時間序列乃[«]，乃|>]，乃问，力间是等同於第二 圖中原本pts方法所產生的離散時間序列XiM，X2[w]，Χ3Μ， 々Μ。類似地，運用對稱性質，可得 其中/7 = 0，1，."，(#/4)-1，而相位旋轉參數忌則如第十一圖 所示。由第十圖可看出，當忌為+1，―丨，+j或_」時，本發明 _ 總共需要(M2)log2#個複數乘法運算，所需記憶體為#個 單元。 第十二圖為本發明與其他三種PTS實現方法的運算量 與s己憶體需求之比較，其中M=4，N的大小分別為64、 256、1024和2048。可以看出，運算量與記憶想需求皆 隨著N的遞增而增加。相較於原來及Kang、.與— 之PTS實現方法’本發明與之pts實現方法中， 17 200539601 所需的乘_：為最少，賴體需求單元也是最少。兩者所 需的乘法數相同，分別為192、1024、5120和11264· 記憶體需求單元也是相同，分別為64、256、1024和 2〇48。惟，本發明之PTS實現方法中，*需將OFDM的 信號長度_，因此仍財GFDM縣具有的特性和優 綜上所述，本發明利用PTS交錯劃分區塊的特性， 且只需用到一個腳FT，提供了一種有效降低正交分頻 多工信號峰均值的方法與裝置。大幅降低運算量，所 需複數乘法數為(7V/2)log2#，所需要的記憶趙為#個單元， 甚且仍保有OFDM原本具有的特性和優點。 惟，以上所述者，僅為本發明之較佳實施例而已，當 不能以此限定本發明實施之範圍。即大凡依本發明申請: 利範圍所作之均等變化與修飾，皆應仍屬本發明專利涵蓋 之範圍内。 200539601 【圖式簡單說明】 第-圖說明_個傳鮮餘通n財關0FDM發射機 的系統方塊圖。 第二圖說明一種使用PTS方法解決OFDM發射端信號之 高PAPR的技術。 第三圖說明將輸入資料义[幻在頻域上劃分為子區塊(或組) 之交錯、鄰接和不規則的三種方式。 第四圖說明Kang、Kim與J〇〇提出之降低pTs運算量的實 現方法。 第五圖說明S amsung公司提出之降低PTS運算量與記憶體 的實現方法。 第六圖為本發明之降低OFDM信號PAPR值的方法的 示意圖。 第七圖說明本發明之降低OFDM信號PAPR值的裝置 的一個結構示意圖。 第八圖為根據第七圖之本發明的第一較佳實施例。 第九圖說明M=2時，第八圖中的相位旋轉參數之設定。 第十圖為根據第七圖之本發明的第二較佳實施例。 第十一圖說明M=4時，第十圖中的相位旋轉參數之設定。 第十二圖為本發明與其他三種PTS實現方法的複數乘法運 算量與記憶體需求之比較。 圖號說明： 601 #點反向快速傅立葉轉換 200539601 603時域信號分割 605組合器 607相位最佳化處理和合成 701解多工器 705記憶體 803組合器 1003組合器 703組合器 707加法器 803a、803b加法器A similar concept was proposed by Samsung in US patent application US 2003/0067866, as shown in the fifth figure. Different from the previous method, each i: the sub-block does not repeat after Z_IFFT, and is directly multiplied by the complex coefficient of the Z point in the time domain to make the time-domain sub-block data orthogonal to each other, so as to facilitate the separation of the receiving end. Data of each sub-block. Since each time-domain sub-block has only Z points, its PAPR value is low, so the PAPR value of the transmitted signal obtained after the phase rotation is added up is also low. Although the number of multiplications required by this method is M.00k) g2L + JV, and the required memory is # units, this method shortens the original OFDM transmission signal of length # to a transmission signal of length jL, which means As a result, the ability of this system to counteract the effects of multi-path channels will be greatly reduced. And in many cases, it is simply impossible to design a complex coefficient multiplier of point I to make the time-domain sub-block data formed by the transmitting end to each other. 200539601 The true father will make it difficult for the receiving end to restore the original transmitted data. [Summary of the Invention] In order to overcome the shortcomings of the traditional PTS implementation method for reducing the PAPR value of the traditional OFDM transmitting end, the main purpose of the present invention is to provide a method and device for reducing the PAPR value of an OFDM signal. This method utilizes the characteristics of pTs to divide blocks in a divided manner, and directly divides a discrete time series of length iV into multiple non-overlapping sub-blocks in the time domain, and then transforms and combines and optimizes a phase. Process and synthesize a complete #point signal pair β, where # is the length of an orthogonal frequency division multiplexed signal and "= 〇, 1, ..., ΛΜ 〇 The present invention only needs to use a 7V4FFT, which can greatly reduce the amount of calculation The required number of complex multiplications is (M2) log2 #, and the required memory is # units. More importantly, the present invention still retains the ability of OFDM to resist multipath channel effects. This method further includes the following steps: First, a discrete time series pair of integer 1) is divided into M subblocks (or groups) that do not overlap (disjoint). The length of each subblock is AT / #, and M is 2. Power (W) and W / M is an integer greater than 1. Then, a combiner (combiner) is used to synthesize this M non-overlapping sub-blocks into another μ sequence of am, where ^ = 1 , 2, ..., λ / and „= 〇, 1,…,. Finally, using the symmetry property, the μ sequences% b] are rotated and added together through phase 12 200539601 to form a complete TV point transmission signal 3f [n]. In the method, the first and second preferred embodiments use M = 2 and M respectively to describe the steps of the time domain implementation method described above. Accordingly, the device for reducing the pAPR value of the OFF signal of the present invention mainly includes a # -point inverse fast Fourier transform, a de-multiplexer, a combiner, a group of memories, and an adder. The #point inverse fast Fourier transform transforms the input frequency domain signal; the transformation parameter becomes the time domain signal χ [«], and the demultiplexer divides X into adjacent sub-blocks that do not overlap each other and have the same length. The combiner synthesizes this from a non-overlapping sub-block into another M lengths; the discrete-time sequence of V / M "ηΙΙΙ. Using the symmetry property, the M sequences of eight [„] are rotated by phase and then added by an adder to form a complete #point signal. Here, the M non-overlapping and equal length sub-blocks and slave sequences Nai ["] is stored in this group of memory. In the case of M = 4, the operation settings of the present invention and the other three PTS implementation methods are compared with the demand of 5 Jiyi. The results show that compared with the original and the PTS implementation methods of Kang, Kim, and Joo, the present invention and Samsung's PTS implementation method require the same number of multiplications and the minimum, and their memory demand units are the same and the minimum. However, the present invention does not need to shorten the signal length of the OFDM, so the original characteristics and advantages of the OFDM system are still maintained. 13 200539601 A detailed description of the deletion and implementation examples and the patent scope of the effective distribution, which will detail the above and other objects and advantages of the present invention. [Embodiment] The sixth figure is a schematic diagram of a method for reducing the PAPR value of an OFFDM signal according to the present invention. First, 'frequency-domain signal of length #_ do #_IFFT 6〇1 operation to obtain the time-domain signal system of length TV, and then divide the system into M non-overlapping sub-blocks according to the adjacent method, as shown in step 60 As shown, the length of the parent and child blocks is AWl /, M is a power of 2, and eight / slave is an integer greater than 1. Then, the M non-overlapping sub-blocks form M discrete time series% [[]] of length # via the combiner 605. Finally, the M discrete time series are subject to steps. The phase of 〇07 is rotated and added together to form a complete #point transmission signal pair ⑴: Φ] = bxyx [n] + b2y2 [n] ^-^ bMyM [n] ⑹ Since the sequence is out of sequence [«] has The symmetry property, so the combiner 605 only needs to form a sequence of length cut ^], h [i], ... aKAWH]}. Based on this, the seventh figure illustrates the method of reducing the PAPR value of the OFDM signal according to the present invention. A structure diagram of the device. This device for reducing the PAPR value of an OFDM signal mainly includes an IFFT 601, a demultiplexer 70, a combiner 703, a group of memories 705, and an adder 707. The MIFFT 601 will input the Frequency domain signal meaning [converted into time domain signal pair ^, 200539601 Demultiplexer 701 divides jcM into M non-overlapping sub-blocks of equal length according to adjacency. The combiner 703 does not overlap each other. The sub-blocks of M are combined into M sequence% [«] of length # / M, where ^ = 1, 2, ..., M and" = 〇, 1, " ·, (Λ "/ M) is smaller than this M sequences w [w] after phase rotation, and is added by the adder 707 to form a length of AMI / transmission signal {? [〇], pair 1], ... ·, Pair (# / Μ) -1]}. By changing different phase rotation parameters &, this sequence of M numbers can be used again] to obtain the next segment transmission signal {? [# / #], Pair ( # / M) + i], ..., for (π / Μ) -ΐ]}. In this way, all the transmission signals are obtained at the last inch? [«]. Among them, M subblocks that do not overlap each other and Μ _ sequences of eight [«] are stored in the memory 705 successively. Take M == 2 as an example for illustration, as shown in the eighth figure. The sequence is especially asked, w == 〇, 1, ... ΛM is demultiplexed by multiplexer 701 into two sub-blocks of length # / 2 {4〇], 41], ..., X [(7V72)-1]} and {χ [ΛΓ / 2], for (7V72 ) +1], ..., X [7V- 'is stored in the memory 705. The combiner 803 combines these two subblocks of length # / 2 to form the following two discrete time series of length #: 馨 yi [^] = x [n] + x [((n-Ν / 2)) ν] [^] = x [n]-x [((n-N / 2)) n] '(7) where ((·) V represents a circular shift (circular shift) # points, w = 〇, i, ..., (# / 2) iota. 8〇3 combination structure is quite simple, two adders, ⑼ known and 803b, to achieve the first (7), two 2_IFFT same structure. It can be shown that the discrete time series obtained in the month type is the same as the discrete time series, which is the same as the discrete time series generated by the original PTS method in the first step. From Equation (7), 15 200539601 [((«+ f» N 1 = [((w -f)) N] ^ x [((n ~ N / 2)) n] + x (n) = ^ [n] less 2 [((”+ f)) n] = h [((”-f)) ”= x [((” — 宁)) " 卜 刺 = —a 间 (8) In other words,乃 [I] = less 1 [〇], less 1 [I + U = less 1 [1], ...,% [iV -1] =% [I-1] h [number] = [〇], h [ f + 1] = -less 2 [l], ···, less 2 [at one η = one h [quan—η (9) So the combiner 803 only needs to form a sequence of length # / 2 (^ [ο] ,% [Η, Nai [(# / 2) -⑴ and Bu Rou Nai, ... ,, Nai, /])-1]}. Since {χ [0], χ [1], " ·, Χ [#-1]} is no longer needed, so memory 705 can be released for {^ 肌 乃 ⑴, -1]} and [y2 [〇], j; 2 [i], less 2 [ (# / 2)-1]} use. In other words, the required memory is # units. Finally, from the formula and formula, we can get x [n ^ f \ ^ bxyx [ri \ -b2y2 [n] (1 〇) where "= 0, 1, ", (M2)-Bu will further express the formula (10) as coffee + hand] P = 〇, ΐ," = 0, 1, ... 1 (11) The phase rotation parameters and \ are shown in the ninth figure. From the eighth figure, it can be seen that 'When & is +1, -1, + j or -j, The total amount of multiplication required by the invention comes from #_IFFT, that is, (M2) log2 # complex multiplication operations are required, and the required memory is # units. When M = 4, the implementation diagram is as shown in the tenth figure The sequence: φ?], W = 0, 1, ..., Α is divided into four sub-blocks of length Λ "/ 4 by the demultiplexer 701, {χ [0], ..., χ [ (Λ // 4) _ 1]}, {χ [# 200539601 /4]"."x[(iW2)- 1]}, {x [TW2], .. "x [(37W4) · 1]} , And {x [3iV / 4], ···, χ [Λ ^ 1]}. The combiner 1003 combines these four subblocks of length τν / 4 to form the following discrete time series of length #: y } [η] = x [n] +-N / 2)) n] + x [((n ~ N / 4)) n] + x [((n-3N / 4)) n] y3 [n] = x [n] + x [((n-N / 2)) n]-x [((n ~ N / 4)) n]-x [((n-37 ^ / 4)) ^] y2 [ n] = x [n]-x [((w-N / 2)) n] + jx [((n-N / 4)) n]-jx [({n-37V / 4)) ^] yA [n] = x [n]-x [((n-N / 2)) n]-jx [((n-N / 4)) n] + jx [((n-3Λ ^ / 4)) ^ (12) The combiner 1003 uses eight adders and an imaginary multiplier to implement formula (I2). Its structure is the same as 4_IFFT. It can be proved that the discrete time series obtained by Φ in (u) is [«], is | >], but the force is equivalent to the discrete time series XiM, X2 [w generated by the original pts method in the second figure. ], × 3Μ, ΜΜ. Similarly, using the symmetry property, we can get / 7 = 0, 1,. &Quot;,(# / 4) -1, and the phase rotation parameter is as shown in the eleventh figure. It can be seen from the tenth figure that when the bogey is +1, 丨, + j or _ ", the present invention requires a total of (M2) log2 # complex multiplication operations, and the required memory is #units. The twelfth figure is a comparison of the calculation amount of the present invention and the other three PTS implementation methods with the memory requirements of S, where M = 4, and the sizes of N are 64, 256, 1024, and 2048, respectively. It can be seen that the amount of computation and memory requirements increase as N increases. Compared with the original and Kang,., And-PTS implementation method, the present invention and the PTS implementation method, 17 200539601 required multiplication _: is the least, and the body demand unit is also the least. The number of multiplications required is the same, which is 192, 1024, 5120, and 11264. The memory requirements are the same, which are 64, 256, 1024, and 2048. However, in the method for implementing PTS of the present invention, the signal length of OFDM needs to be _, so it still has the characteristics and advantages of GFDM county. As described above, the present invention utilizes the characteristics of PTS to divide blocks, and only needs to use One foot FT provides a method and device for effectively reducing the peak-to-average value of the orthogonal frequency division multiplexed signal. Significantly reduce the amount of operation, the complex multiplier required is (7V / 2) log2 #, the required memory is # units, and the original characteristics and advantages of OFDM are still retained. However, the above are only preferred embodiments of the present invention, and the scope of implementation of the present invention cannot be limited by this. That is to say, all equal changes and modifications made in accordance with the present invention by the scope of the invention shall still fall within the scope of the patent of the present invention. 200539601 [Brief description of the diagram] Figure-Description of the system block diagram of a Chuanxianyutongn Caiguan 0FDM transmitter. The second figure illustrates a technique that uses the PTS method to solve the high PAPR of the OFDM transmitting signal. The third figure illustrates the three ways of interleaving, contiguous and irregular dividing the input data into subblocks (or groups) in the frequency domain. The fourth figure illustrates the implementation method proposed by Kang, Kim and Joo to reduce the amount of pTs operation. The fifth figure illustrates the implementation method proposed by Samsung to reduce the amount of PTS operations and memory. The sixth figure is a schematic diagram of a method for reducing a PAPR value of an OFDM signal according to the present invention. The seventh figure illustrates a structure diagram of the apparatus for reducing the PAPR value of an OFDM signal according to the present invention. The eighth figure is a first preferred embodiment of the present invention according to the seventh figure. The ninth figure illustrates the setting of the phase rotation parameter in the eighth figure when M = 2. The tenth figure is a second preferred embodiment of the present invention according to the seventh figure. The eleventh figure illustrates the setting of the phase rotation parameter in the tenth figure when M = 4. The twelfth figure is a comparison of the complex multiplication operation amount and memory requirements of the present invention and the other three PTS implementation methods. Drawing number description: 601 #point inverse fast Fourier transform 200539601 603 time domain signal segmentation 605 combiner 607 phase optimization processing and synthesis 701 demultiplexer 705 memory 803 combiner 1003 combiner 703 combiner 707 adder 803a 803b adder
2020