201012136 ' 六、發明說明： 【發明所屬之技術領域】 本發明係關於一種數位視訊廣播系統，特別是有關一種利用離散傅立 葉變換運算來估計通道之視訊廣播系統及其通道估計方法。 【先前技術】 數位視訊廣播(Digital Video Broadcasting，以下簡稱為DVB)是國際上 承認用於數位電視之標準規格。1990年中旬被開發出來的地面數位視訊廣 ❹播(Digital Video Broadcasting-Terrestria卜以下簡稱為DVB-T)具有可攜帶以 及固疋接收之特性’係著重考ΐ接收器之開發價格而設計的。為了實現機 帶以及移動接收，DVB-T使用2個天線之分集接收技術，從而可在惡劣之 情況下也能實現高速移動接收。然而在DVB-T之移動性實驗令，出現不能 適用於其他移動多媒體應用服務等問題，發現DVB-T用作手機廣播是有缺 點的’因此制定了用於攜帶式機器新的DVB標準，即手持式數位視訊廣播 (Digital Video Broadcasting-Handheld，以下簡稱為 DVB-Η) 〇 DVB-Η 係考量 可攜帶以及透過電池來工作之性質而制定的’但由於是根據DVB-T之標 β 準，因此大部分可與DVB-T互換。 DVB係利用正交分頻多工技術(Qj^ogonai FreqUenCy Diyisi〇n Multiplexing ’以下簡稱為OFDM)將符號、導頻以及傳輸參數信號 (TransmissionParameter Signaling，以下簡稱為TPS)分配後傳送至各副載波 (Sub-Carrier)。導頻由全部〇fdM符號被傳送到固定位置之連續導頻以及每 4個OFDM符號位置發生週期性變化之分散導頻構成。xps將現在接收的 信號傳送方法所相關之情報透過68個OFDM符號(1幀)傳送。TPS在復原 時以不需要通道情報的差動二相位移鍵(Differential Binary Phase Shift Keying，以下簡稱為DBPSK)方式傳送，但數位信號係以正交位移鍵 201012136 ‘ (Quadrature Phase Shift Keyipg，以下簡稱為 QPSK)、16-正交振幅調變 (16-Quadrature Amplitude Modulation，以下.簡稱為 16-QAM)以及 64-正交振 幅調變(64-Quadrature Amplitude Modulation ’ 以下簡稱為 64_qam)其中一種 方式調變後傳送，因此需要準確之通道情報且所估計之通道準確度與整個 系統之性能有密切的關係。在DVB中可用於通道估計之通道情報包括從已 知位置傳送已知值的分散導頻和連續導頻。 請參閱第1圖，係繪示DVB系統中，1個OFDM符號所包含之導頻之 大致形狀。如第1圖所示’在連續導頻的情況下，全部OFDM符號連續傳 〇 送至相同的副載波位置上。但是連續導頻的數量與整個副載波數量相較之 下過小’且導頻之間的間距不規則。因此通道估計中只使用連續導頻是不 夠的，必須同時再使用每4個OFDM符號傳送至發生週期性變化之副載波 位置的分散導頻。 在DVB系統中，一般的通道估計方法為導頻所在位置的通道估計、含 有分散導頻之副載波的符號座標轴插值、以及以1個0FDM符號為單位而 工作之頻率軸插值。第1圖之〇FDM符號中存在連續導頻和分散導頻，第 n個0FDM符號的第k個副載波所包含的導頻Υη*如公式1表示。 〇 【公式1】201012136 ' VI. Description of the Invention: [Technical Field] The present invention relates to a digital video broadcasting system, and more particularly to a video broadcasting system for estimating a channel by using a discrete Fourier transform operation and a channel estimating method thereof. [Prior Art] Digital Video Broadcasting (hereinafter referred to as DVB) is an internationally recognized standard specification for digital television. Developed in mid-1990, Digital Video Broadcasting-Terrestria (hereinafter referred to as DVB-T) has the characteristics of portability and solid-state reception. It is designed to focus on the development price of the receiver. In order to realize the band and mobile reception, DVB-T uses a diversity reception technology of two antennas, so that high-speed mobile reception can be realized under severe conditions. However, in the DVB-T mobility experiment, there are problems that cannot be applied to other mobile multimedia application services, and it is found that DVB-T is used as a mobile phone broadcast. Therefore, a new DVB standard for portable devices has been developed. Handheld digital video broadcasting (Digital Video Broadcasting-Handheld, hereinafter referred to as DVB-Η) 〇 DVB-Η is based on the nature of portable and battery-operated work, but because it is based on the DVB-T standard, Therefore most can be interchanged with DVB-T. The DVB system uses the orthogonal frequency division multiplexing technique (Qj^ogonai FreqUenCy Diyisi〇n Multiplexing' hereinafter referred to as OFDM) to allocate symbols, pilots, and transmission parameter signals (TPS) to each subcarrier. (Sub-Carrier). The pilot consists of contiguous pilots that are transmitted to all fixed FDD symbols to a fixed position and scattered pilots that periodically change every 4 OFDM symbol positions. The xps transmits the information related to the currently received signal transmission method through 68 OFDM symbols (1 frame). The TPS is transmitted in the form of a Differential Binary Phase Shift Keying (DBPSK) that does not require channel information during recovery, but the digital signal is orthogonally shifted by 201012136 ' (Quadrature Phase Shift Keyipg, hereinafter referred to as One of the modes of QPSK), 16-Quadrature Amplitude Modulation (hereinafter, referred to as 16-QAM for short) and 64-Quadrature Amplitude Modulation (hereinafter referred to as 64_qam) Transmitted, so accurate channel intelligence is required and the estimated channel accuracy is closely related to the performance of the overall system. Channel information available for channel estimation in DVB includes scattered pilots and continual pilots that transmit known values from known locations. Referring to Figure 1, the general shape of the pilot included in one OFDM symbol in the DVB system is shown. As shown in Fig. 1, in the case of continual pilots, all OFDM symbols are continuously transmitted to the same subcarrier position. However, the number of consecutive pilots is too small compared to the number of entire subcarriers' and the spacing between pilots is irregular. Therefore, it is not sufficient to use only continuous pilots in the channel estimation, and every 4 OFDM symbols must be used simultaneously to transmit to the scattered pilots of the periodically changing subcarrier positions. In the DVB system, the general channel estimation method is channel estimation at the position where the pilot is located, symbol coordinate axis interpolation of subcarriers including scattered pilots, and frequency axis interpolation operating in units of 1 OFDM symbol. In Fig. 1, there are contiguous pilots and scattered pilots in the FDM symbols, and the pilot Υη* included in the kth subcarrier of the nth OFDM symbol is expressed by Equation 1. 〇 [Formula 1]
Yn,k = Pn,k -Hn k +Nn>k 其中，Pn,k表示該位置的導頻，Hn，k為其導頻所通過的通道值，而风太為 加法性白色尚斯雜訊(Additive White Gaussian Noise，以下簡稱為AWGN)， AWGN將會被加到通過通道的值。 由於接收端提前知道導頻所在位置及其值，因此可根據公式2估計出 第η個〇FDM符號中第k導頻所在部分之通道值。 【公式2】 会砧=亡，(Y>a = pn>k. + NnJt) 201012136 、 所估計之通道準確度與增加的awgn的大小成反比，與導頻的大小成 正比。DVB系統為了透過導頻來提高估計之通道準確度，將比數位信號的 平均值大於4/3的值分配給導頻。 當對所有導頻執行上述過程時，將可知道第i圖中之導頻所在位置(黑 點)下的所有通道值。 對含有分散導頻的副載波進行符號座標軸插值的通道估計階段中，分 散導頻並不存在於相同副載波的全部〇FDM符號，而是以4個符號為間距 而存在，因此利用分散導頻’將每隔4個符號存在相同的副載波所相應的 © 通道值’進行符餘縣插值’從雜計丨其職波巾林在分散導頻的 OFDM符麟減之值。請參㈣2圖，鱗示執行賊座標抽插值 時，可知道通道值的位置。執行該通道估計方法時，能得知每隔3個副載 波的估計通道值，可將此利用到全部OFDM符號之相同的頻率軸插值。 頻率轴插值後對所有副載波所相應的通道值進行估計的階段，在利用 分散導頻來估計通道的OFDM系統中，一般係利用1次線性插值或2次以上複 雜插值、低通濾波器(LPF)等執行頻率軸插值。 然而’ DVB系統需要保障在所謂單頻網(single Frequency Network，以 φ 下簡稱為SFN)的特殊通道中能流暢地接收資料，因此很難使用普遍的頻率 轴插值方法。其原因在於SFN環境為了在OFDM中適應多種路徑的通道環 境，包括各種長度的多種路徑通道環境，其中包括所使用的循環字首(Cyclic Prefix)的80%以上具有通造延遲的通道(SFN Long Channel)。使用1次插值 的情況下，在插值的過程中會除去延遲較長的通道成份，因此無法得到較 佳性能。此外，即使使用低通濾波器，根據通道延遲長度而需要多個分別 具有不同截止頻率(CutoffFrequency)的濾波器’用於掌握通道延遲程度的工 作。 【發明内容】 201012136 ' 本發明之一目的在於提供一種DVB系統及其通道估計方法，係利用離 散傅立葉變換(Discrete Fourier Transform，以下簡稱為DFT)估計出通道 為了達到上述目的，根據本發明之DVB系統包括：一反離散傅立葉變 換(Inverse Discrete Fourier Transform，以下簡稱為 IDFT)運算部、—通道耽 衝響應估計裝置、一 DFT運算部以及一第一乘法器。IDFT運算部根據所接 收之DVB信號而進行符號座標軸插值時，對以3個副載波為間隔而存在之 通道情報進行IDFT運算。通道脈衝響應估計裝置根據idft運算部之輪 出，設置一視窗用以判斷通道脈衝響應。DFT運算部對視窗内之!ορρ運算 ® 部之輸出進行DFT運算。第一乘法器將DFT運算部之輸出乘以3倍以調整 整個能量比率。 根據本發明之實施例，通道脈衝響應估計裝置包括一功率值平均計算 部、一第二乘法器以及一剪切視窗發生部。功率值平均計算部將^FT運算 之功率值平均成與與一循環字首之長度相同，其中循環字首係由DVB系統 根據正父分頻多工技術符號長度所選擇。第二乘法器將與循環字首之長度 相等之IDFT運算之平均功率值乘以一 α值，藉此設置一臨界值，其中“值 係根據DVB系統之一資料調變方式而定。剪切視窗發生部將運算結 〇 果之功率值超出臨界值之區間設置成視窗。 根據本發明之一實施例’剪切視窗發生部在運算結果之功率值超 出臨界值之區間追加用於增加判斷為通道脈衝響應部分之前後部分之預留 長度’並設置視窗。 根據本發明之另一實施例，α值根據資料調變方式之位元錯誤率以及 信號雜訊比被設置成不同值。 根據本發明之又-實施例，DVB系統更包括一增益增強部，用於補償 由IDFT運算結果而估計之通道邊緣之增益損失。 根據本發明之DVB系統之通道估計方法包括：根據所接收之一 DyB k號而進行符號座標軸插值時’對以3個副載波為間隔而存在之通道情報 201012136 -進行腑運算；根據IDFT運算之-輪出，設置一視窗用以綱通道脈衝 0應’對視固内咖丁運算之輸ώ進行DFT運算；以及將DFT運算之輪出 乘以3倍以調整整個能量比率。 、，根據本利之—實細’設置視窗之步魅括：將膝^運算之功率值 ，成與-循環字首之長度相同，其中循環字首係由DVB系統根據〇職 符號長度所選擇；將與循環字首之長度相等之聊丁運算之平均功率值乘以 一α值，藉此設置—臨界值’其中碰係根據dvb祕之—資料調變方式 而疋’ 將IDFT運算結果之功率值超出臨界值之關設置為視窗。 ❿ 根據本發明之另—實補’設置視窗的階段還包括：在IDFT運算結果 之功率值超出臨界值之區間增加預留長度的階段，其中預留長度係用於增 加判斷為通道脈衝響應部分之前後部分。 根據本發明之X-實補’ DVB祕之通道糾方法更包括：對根據 DFT運算科而估狀通道絲之增益損失進行麵。 本發明的優點在於：為了透過利用DVB系統中使用的導頻所表現出 來的IDFT結果的特徵，從而得出使用於通道估計的通道脈衝響應成分， 利用根據資料調變方式而發生變化的臨界值和具有有利長度的多餘空 ® 間，由此設置剪切視窗，並對所得到的通道脈衝響應進行DFT運算而估 s十通道。為了提高估計的通道的邊緣部分的估計準確度而進行增益補償， 由此改善通道估計性能。如此提高的通道估計的準確度，在採用使用較高 編碼速率的QAM方式時提供更大的增益。 【實施方式】 以下將參照附圖就本發明的具體實施例進行詳細說明，各附圖中相同 符號係表示相同的元件。 請參閱第3圖，係繪示本發明之一實施例之DVB系統3〇〇cDVB系統 300包括一第一DFT運算部3〇2、一時間軸插值部（Time Axis 201012136 _ 、一 IDFT運算部306、一通道脈衝響應估計裝置308、一 視窗發生部310、一第二DFT運算部312、一第一乘法器314、一符號延遲 部(Symbol Delay for Time-Axis Interpolation^ 16、一增益增強部 318 以及一 解映射器(Demapper)320。 第一 DFT運算部302接收一 DVB信號後，執行將時域(Time Domain) 數位信號轉換為頻域(Frequency Domain)數位信號之運算。 時間軸插值部304對第1圖中的黑點所相應的分散導頻進行時間軸插 值，進而估計出有嵌入資料部分(白點)的通道值。對第1圖中之黑點進行時 φ 間軸插值後變成如第2圖所示之形狀。 IDFT運算部306為DFT運算的逆過程，執行將頻域數位信號轉換為時 域數位彳s號之運算，IDFT部306之輸出如第4圖所示。 通道脈衝響應估計裝置308從BDFT部306之輸出中找出實際之通道脈 衝響應(Channel Impulse Response)的部分。 視囪發生部310從IDFT部306之輸出中，將判斷為通道脈衝響應 (Chamiel Impulse Response ’ CIR)的部分上設置相應的視窗，所謂設置相應 的視窗(Windowing)係指在第4圖中僅留下實際通道脈衝響應，除去不必要 0 的部分。 第二DFT運算部312執行將視窗發生部训之時域輸出轉換為頻域輸 出之運算。 第-乘法器314將第二DFT部312之輸出乘以3倍。在第2圖中，每 隔3個副載波(Sub-Carrier)可知道一個通道情報。對這樣 丽運算箱運算時，可以得到如第4圖所示之相同的通道脈 複3次的雜。其巾實際通道脈衝響應為3辦之丨個，因此其餘2個重 複的通道脈衝響應會被删除，此時能量只剩下整_ 1/3，故在通過第二 DFT運舁部312後重新乘以3來調整能量比率。 符號延遲部耵6係為了調整現在所接收的〇FDM符號之被估計的通道 201012136Yn,k = Pn,k -Hn k +Nn>k where Pn,k represents the pilot at this position, Hn,k is the channel value through which the pilot passes, and the wind is too additive white white noise (Additive White Gaussian Noise, hereinafter referred to as AWGN), AWGN will be added to the value of the pass channel. Since the receiving end knows in advance the location of the pilot and its value, the channel value of the portion of the nth 〇FDM symbol in which the kth pilot is located can be estimated according to Equation 2. [Formula 2] Anvil = Death, (Y>a = pn>k. + NnJt) 201012136 The estimated channel accuracy is inversely proportional to the magnitude of the increased awgn and is proportional to the size of the pilot. In order to improve the estimated channel accuracy by using pilots, the DVB system assigns a value greater than 4/3 of the average value of the digital signal to the pilot. When the above procedure is performed for all pilots, all channel values under the pilot position (black point) in the i-th picture will be known. In the channel estimation phase of symbol coordinate axis interpolation for subcarriers with scattered pilots, the scattered pilots do not exist in all 〇FDM symbols of the same subcarrier, but exist at intervals of 4 symbols, thus utilizing scattered pilots 'There will be the value of the corresponding channel value of the same subcarrier every 4 symbols', and the value of the OFDM symbol of the scattered pilot is subtracted from the miscellaneous. Please refer to (4) 2, the scale shows the position of the channel value when the thief coordinates are inserted. When the channel estimation method is performed, the estimated channel values for every three subcarriers can be known, and this can be interpolated using the same frequency axis of all OFDM symbols. In the stage of estimating the channel value of all subcarriers after frequency axis interpolation, in the OFDM system using the scattered pilot to estimate the channel, generally, one linear interpolation or more complex interpolation and low-pass filter are used ( LPF) and so on perform frequency axis interpolation. However, the DVB system needs to securely receive data in a special channel called a single frequency network (hereinafter referred to as SFN), so it is difficult to use a common frequency axis interpolation method. The reason is that the SFN environment is adapted to accommodate multiple path environments in OFDM, including multiple path channel environments of various lengths, including more than 80% of the Cyclic Prefix used to have a delay-producing channel (SFN Long). Channel). In the case of one interpolation, the channel components with a long delay are removed during the interpolation process, so that better performance cannot be obtained. Further, even if a low-pass filter is used, a plurality of filters respectively having different cutoff frequencies (Cutoff Frequency) are required for grasping the degree of channel delay depending on the channel delay length. SUMMARY OF THE INVENTION 201012136 'An object of the present invention is to provide a DVB system and a channel estimation method thereof, which utilizes Discrete Fourier Transform (hereinafter referred to as DFT) to estimate a channel in order to achieve the above object, and DVB according to the present invention. The system includes an Inverse Discrete Fourier Transform (hereinafter referred to as IDFT) operation unit, a channel buffer response estimating device, a DFT operation unit, and a first multiplier. When the IDFT calculation unit performs symbol coordinate axis interpolation based on the received DVB signal, the IDFT calculation unit performs IDFT calculation on the channel information existing at intervals of three subcarriers. The channel impulse response estimating means sets a window for judging the channel impulse response based on the rotation of the idft arithmetic unit. The DFT calculation unit performs DFT calculation on the output of the !ορρ operation ® part in the window. The first multiplier multiplies the output of the DFT operation unit by a factor of three to adjust the overall energy ratio. According to an embodiment of the present invention, the channel impulse response estimating means includes a power value averaging calculating portion, a second multiplier, and a cut window generating portion. The power value averaging calculation unit averages the power values of the ^FT operation to be the same as the length of a cyclic prefix, wherein the cyclic prefix is selected by the DVB system according to the length of the positive-parity multiplex technique symbol. The second multiplier multiplies the average power value of the IDFT operation equal to the length of the cyclic prefix by an alpha value, thereby setting a threshold value, wherein "the value is determined according to a data modulation mode of the DVB system. The window generating unit sets a section in which the power value of the operation result exceeds the critical value as a window. According to an embodiment of the present invention, the clipping window generating unit adds a value for increasing the value of the calculation result to a value exceeding a critical value. The channel impulse response portion has a reserved length before and after the portion and sets a window. According to another embodiment of the present invention, the alpha value is set to a different value according to the bit error rate of the data modulation mode and the signal noise ratio. According to still another embodiment of the invention, the DVB system further includes a gain enhancement unit for compensating for gain loss of the channel edge estimated by the result of the IDFT operation. The channel estimation method of the DVB system according to the present invention includes: according to one of the received DyB When the symbol coordinate axis is interpolated with the k number, 'the channel information 201012136 that exists at intervals of 3 subcarriers is used for the calculation; according to the IDFT operation - Out, set a window for the channel pulse 0 should be 'DFT operation on the input of the internal calculus operation; and multiply the round of the DFT operation by 3 times to adjust the overall energy ratio. The actual detail of the setting window is as follows: the power value of the knee ^ operation is the same as the length of the - loop word prefix, wherein the cycle word is selected by the DVB system according to the length of the job symbol; The average power value of the equal length of the chat operation is multiplied by an alpha value, thereby setting - the critical value 'where the touch is based on the dvb secret - the data modulation method 疋 'the power value of the IDFT operation result exceeds the critical value It is set as a window. ❿ The stage of setting the window according to another embodiment of the present invention further includes: a stage of increasing the reserved length in a range in which the power value of the IDFT operation exceeds a critical value, wherein the reserved length is used to increase the judgment as The channel impulse response portion is before and after the portion. The X-real complement DVB secret channel correction method according to the present invention further comprises: performing surface estimation of the gain loss of the channel filament according to the DFT operation section. In order to obtain the channel impulse response component used for channel estimation by using the characteristics of the IDFT result expressed by the pilot used in the DVB system, the threshold value and the advantageous length which are changed according to the data modulation method are used. The excess space is set, and the clipping window is set, and the obtained channel impulse response is subjected to DFT operation to estimate s ten channels. Gain compensation is performed to improve the estimation accuracy of the estimated edge portion of the channel, thereby improving Channel estimation performance. The accuracy of the channel estimation thus increased provides greater gain when using the QAM method using a higher coding rate. [Embodiment] Hereinafter, a specific embodiment of the present invention will be described in detail with reference to the accompanying drawings. The same reference numerals are used to refer to the same elements in the drawings. Referring to Figure 3, a DVB system according to an embodiment of the present invention includes a first DFT operation unit 3, a time axis. Interpolation unit (Time Axis 201012136 _, an IDFT calculation unit 306, one-channel impulse response estimation device 308, a window generation unit 310, and a first DFT computation unit 312, a first multiplier 314, a delay unit symbol (Symbol Delay for Time-Axis Interpolation ^ 16, a reinforcing portion 318, and a gain demapper (Demapper) 320. After receiving the DVB signal, the first DFT operation unit 302 performs an operation of converting a Time Domain digital signal into a frequency domain (Frequency Domain) digital signal. The time axis interpolation unit 304 performs time axis interpolation on the scattered pilot corresponding to the black point in Fig. 1, and further estimates the channel value of the embedded data portion (white point). When the black point in Fig. 1 is performed, the inter-φ axis is interpolated to become the shape shown in Fig. 2. The IDFT calculation unit 306 performs an operation of converting the frequency domain digital signal into the time domain digit 彳s number in the inverse of the DFT operation, and the output of the IDFT unit 306 is as shown in Fig. 4. The channel impulse response estimating means 308 finds the portion of the actual channel Impulse Response from the output of the BDFT section 306. The view generating unit 310 sets a corresponding window from the output of the IDFT unit 306, and determines that the channel impulse response (CIRiel Impulse Response 'CIR) is set. The so-called setting of the corresponding window (Windowing) means only in FIG. Leave the actual channel impulse response, removing the portion of unnecessary 0. The second DFT operation unit 312 performs an operation of converting the time domain output of the window generation section into the frequency domain output. The first multiplier 314 multiplies the output of the second DFT section 312 by a factor of three. In Figure 2, one channel intelligence is known for every three subcarriers (Sub-Carrier). When such a calculation box is operated, the same channel pulse as shown in Fig. 4 can be obtained three times. The actual channel impulse response of the towel is 3, so the other 2 repeated channel impulse responses will be deleted. At this time, only _ 1/3 of the energy is left, so after passing through the second DFT engine 312, Multiply by 3 to adjust the energy ratio. The symbol delay unit 耵6 is for adjusting the estimated channel of the currently received 〇FDM symbol 201012136
" 與資料的同步，而對符號進行延遲〇利用分散導頻來進行時間抽插值時僅 使用一個OFDM符號是不夠的，需要儲存多個OFDM符號後，再利用之前 的符號之通道情報來進行插值，因此會發生OFDM符號的延遲。也就是說 為了估計現在所接收的OFDM符號的通道，還需要多個符號之後的0FDM 符號。 增益增強部318用於補償被估計之通道邊緣部分之增益損失。由於使 用DFT運算在時間軸上將通道脈衝響應設置視窗(windowing)之方式，因此 在頻率軸上出現Sine函數被迴旋(Convolution)的狀態，導致被估計之通道 Ο 邊緣部分之增益出現衰減的現象，邊緣增益增強部318對該現象進行逆補 償0 解映射器320對通過通道後所接收之信號進行通道補償，使其加工成 可用於後續工作中。DVB系統300發送數位資料時，根據需要透過qPSK、 16-QAM以及64-QAM之其中一種方式進行調變。例如在最簡單的qPSK 情況下’從多個平面映射到{1+j，1-j ’ -1+j，中的一個點。 在DVB系統300中’頻率軸插值方法透過IDFT部3〇6、通道脈衝響 應估計裝置308、視窗發生部310、第二DFT運算部312以及增益增強部 Ο 318來實現。頻率轴插值方法係透過符號座標轴插值來對第2圖中以3個副 載波為間距而存在的通道情報進行IDFT運算，藉此得出〇FDM符號通過 時間軸之通道脈衝響應，但該通道脈衝響應在每3個IDFT之輸入中僅有^ 個值，其他則被填充，因此如第4圖所示，表現出通道脈衝響應重複3 次之形狀。其中僅訂第i雛衝，將第2麵第3個脈衝賴的部分填 充，〇,後，再執行DKT運算可以得到頻率軸插值的結果。此時idft運算的 、结果中整個能量的2/3被，〇’填充，因此需要透過第一乘法器314乘以3件 來調整整個能量。 ° 從IDFT運算的結果找出準確的通道脈衝響應的方法將會影響被 的通道的準確度，在DVB系統中利用DFT運笪 异進行頻率軸插值時需要考 201012136 慮如下特性：DVB系統並未使用整個快速傅立葉變換(細⑽細 Transform ’以下簡稱為FFT)運算大小的所有副載波。例如，在2κ的fft 運算的情訂，魏__ 17G5 _贼，其餘職充，Q，讀進行idft 運算再發送信號。由於·端不知道接收部分被填充為，G，的通道情報，因 此利用DFTS算來進行辭軸插值時同樣也只能用，〇,來填充該部分。然而 這與在鮮軸對所估計的通道值罩上長度為窗之矩形視窗㈣加糾肛 Window)具有侧的錄，目此運算之絲憎纽錄騎道脈衝 響應的形狀並不異常’其表現了當副載波間距為丨且以2〇48/17()5的週期來 ® 進行零交點(Zer〇_CroSsing_ Sinc函數被採樣的的狀態。從第4圖中也可以 知道在較大值的脈衝周邊存在重複而較小的脈動成分，由於這些脈動成分 是源於通道脈衝響應，因此為了提高頻率軸插值的準確度，應將這些 部保留。 在剛系、統中利用DFT運算來進行頻率轴插值時如果只考慮上述特 性，最好保留從IDFT運算結果中具有較大值所在位置向前及向後盡可能寬 的部分’但如果健部分過寬就等於捨棄了糊贿運算的 的優點。-般情況下，通道脈械分出現在較窄的部分，而雜訊成分則分 參佈在整個位置，因此利用DFT運算來進行插值時，在非通道脈衝成分的位 置上填充’〇’時具有能除去佔有較寬位置的雜訊成分的效果。 本實施例之通道麟響應倾方法考慮了上述躲，#輯idft運算 結果而計算判斷為通道脈衝成分之臨界值時，利用咖丁運算結果中之一部 分功率和資料調變方式決定預留值，即用於保留判斷為通道脈衝部分的前 後長戽時為透過資料調變方式來決定。 請參閱第5圖’絲示第3圖之通道脈衝響應估計裝置之功能方 塊圖》通道脈衝響應估計裝置308包括一功率值計算部5〇2、一功率值=均 計算部504、—第二乘法器5〇6以及一剪切視窗發生部5〇8。 ' 功率值計算部5〇2用於計算猶部3〇6(如第3圖所示)之輪出功率。 201012136 由於IDFT 3〇6(如第3圖戶斤示)<輸出為複數，因此功率值言十算透過(a 來進行計算。 功率值平均計算部5〇4用於求得與循環字首之長度相等的 IDFT部(如 第3圖所不)3〇6輸出功率聊丁辦如第3圖所示例6輸出功率的平均值。 循環字首之值由發送端決定，當DVB系統3〇〇(如第3騎示)選擇整個 OFDM長度的1/4、1/8、1/16以及1/32之其中一個後進行傳送。 __第一乘法器部506將與循環字首之長度相等之IDFT部306(如第3圖所 不)之輸出辨值的平均值，細__ α值，其巾α鎌翻變方撕舰、 © 16_QAM 或 64_QAM)而決定。 剪切視窗發生部5〇8在確認臨界值(rj^shoid)以及增加預留後，設置剪 切視臨界值為與循料首之長度相等之咖丁運算的功率值的平均值乘 以&後的值’如帛6圖所示之臨界值。喊認臨界值係指在IDFT運算之結果 中找出功率值超出臨界值之部分。增加預留係指在聊丁運算之結果中功率 值超出臨界值之部分的前後部分，保留與第6圖中，G，長度相等的範圍後設 置剪切視窗。 在本實施例中’於執行一定時間之〇FDM符號時間同步後，由於時間 ⑩同步誤差不大，目此在執行時可以假設通道脈衝響應成分都&含於〇FDM 符號之循環字首。如第2圖所示，在符號座標轴插值結束的狀態下對通道 進行DFT運算時，會有相同的通道脈衝響應重複3次從運算結 果的前侧’進行與鄕字首的長度相等的平均功率的計算。 、其次’根據在OFDM符號中各資料所採用之調變方式而取得不同〇[值， 並將平均功率乘以α值作為判斷通道脈衝響應成分之臨界值。其中根據資 料調變方式取得獨α值是由於可得到適當標準的位元錯誤雜社Err〇r BER)以及仏號雜訊比(Signal-to-Noise Ratio，SNR)。簡單地說，以 Q SK調變的資料被傳送時可接收信號的信號雜訊比，小於被64-QAM調 變的資料的可接收信號的信號雜訊比，因此根據傳送方式來取得不同之臨 11 201012136 ' 界值’可得到最佳之性能，將這樣決定的臨界值作為基準來與IDFT運算結 果的各功率進行比較，並找出通道脈衝響應成分的位置。 接著設置剪切視窗(Cut Window) ’剪切視窗係在判斷為通道脈衝響應成 分之各位置的前後保留預留的範圍，並允許屬於該範圍的通過，而不屬於 該範圍的則以’〇,來填充。考慮到在第1次!DFT運算結果中通道脈衝響應重 複了 3次，因此要排除第2次和第3次成分。如前所述，透過預留範圍的 值來保留通道脈衝響應成分的前後部分之理由是為了保留基於脈衝響應的 脈動成分。 〇 根據數位信號的調變方式來取得不同預留範圍的值包括兩個原因：第 疋與α值根據傳送方式而具有不同值的理由相同；第二是QpSK方式和 QAM方紅狀有差異。在QPSK的情盯，卩、減信叙她來判斷其 值’因此對所估計的通道大小不敏感，但在QAM的情況下對信號進行恢復 時需要信號的相位和大小，因此為了減少發生所估計的通道邊緣部分之增 益損失的情況，要取大於QPSK方式的預留值。 對IDFT運算結果設置如上述之視窗後進行啦運算時^制被頻 率轴插值的通道情報。 ® 方面將通道情報以上述方法進行頻率軸插值時，包含了根據調變 方式之適當長度通道脈衝之脈動成分，但為了除去雜訊成分而不得不限制 其長度。所謂限制長絲設置視窗是在時間軸上對通道脈衝響應罩上矩形 視窗(Rectangular Window) ’對其結果進行DFT運算會導致在被估計的通道 的邊緣上發生增益敏(GamLQSS)。為了更簡要地說明增益損失魏請參 閱第7圖係'徐示在僅存在有非常小的AWGN $通道中利肖運算來 對通道進行估計。 在第7圖中’將设置剪切視窗時的預留值設置為ι〇。第7圖最上方的 圖表不利用DFT運算所估計的通道的實數值，中間的圖表示虛數值，最下 方的圖表示所估計的通道的絕對值。在沒有實際通過通道時所估計的通道 12 201012136 -中具有虛數值是因絲將符鱗_步黯_故。要注意的是’被 估計的通道越到邊緣部分與實際通道值的差異越大。為了進一步詳細說 明’請參閱第8圖’係繪示將所估計的通道與實際通道兩者之均方誤差值。 由圖中可知，在邊緣部分所估計的通道之準财非常低，且這樣的 性是由於對IDFT運算的結果設置視窗而產生的。 當使用本實施例的頻率軸插值方法時，在判斷為通道脈衝響應部分之 前後部分__度的前顧，耻対城窗財絲近存在通道 麟響應成分雜HT被絲罩住，赌舰財轉麵舰衝響應成 ®分時，對其進行DFT運算而得到的通道之邊緣部分不會偏向於實數或虛 數’僅具有相差不多的增益損失。若未設置在剪切視窗的中央，而是在偏 向-侧的麟下設魏窗時，通過DFT運算轉_通道邊緣部分的增益 損失值的實數部分和虛數部分會不同，並會被估計成恢復相位時的狀態， 從而想要恢復是非常困難的。 請參閱第9圖’騎示第3圖之增益增強部318之邊緣增益補強方法。 以下將說明本實施例在進行頻率軸插值時，對透過DFT運算後所得到的通 道之邊緣部分進行如第9圖中三角形斜線區域大小之增益損失的補償方法。 ❹ 在第9圖中，[b/(a+b)]之值不會因為受到通道條件或其他影響而發生較 大變化疋幾乎保持疋值，但是c值與上述所使用的煎切視窗的幅度成 反比’其巾f切視窗的幅舰著傳送方式和通道條件而發生變彳卜因此每 個OFDM符號都根據剪切視窗的幅度計算出c值後，對所估計的通道的邊 緣部为以接近於直角三角形的形狀進行補償，從而可提高通道估計的性 能。該方式只對所估計的通道的大小進行準確的調整，因此，使用對通道 大小敏感的QAM方式時，比使用對通道大小相對不敏感的QpSK方式，能 取得更大的增益。 表1為各種通道條件和信號雜訊比下，以本發明之DFT運算方式進行 頻率軸插值’而得到的通道邊緣2〇瞬時值的均方誤差值。表1中的GB 〇ff 13 201012136 表不未使用邊緣增益增強(Gain Boosting)，而GB On則表示使用了增益增 強。從表1中可知，除了在信號雜訊比極惡劣的條件外，使用增益增強時 所估計的通道的準確度與通道條件無相關地有所提高。特別是在未使用增 益增強時，即使信號雜訊比增大，所估計的通道的準確度幾乎沒有增加。 运表不所料的通道絲部分神確度下降原目與健雜減並無關聯。 相反地，增益增強對邊緣部分的增益損失進行了近似值的補償，因此可以 確認信號雜訊比越大，所估計的通道的準確度也越好。 表1 通道形式 (Channel Type) SNR(dB) >pc 1 GBOff GBOn AWGN 5 0.12985 0.14930 10 0.05912 0.01222 15 0.05626 0.00575 20 0.05362 0.00384 Portable 5 0.13490 0.15813 10 0.06017 0.02521 15 0.05232 0.01615 20 0.04941 0.01296 SFNSL 5 0.15176 0.19871 10 0.07726 0.04474 15 0.06421 0.02820 20 0.06018 0.02239 SFNSS 5 0.15708 0.16244 10 0.07609 0.03637 15 0.06762 0.02345 201012136 20 0.06547 0.01953 請參閲第ίο圖以及第u圖，係繪示透過根據頻率軸插值之維特比 (Viterbi)運算後’對位元錯誤率進行電腦模擬之結果。電腦模擬之基本假設 如下：系統模型以DVB系統為基礎，FFT大小為8K模型，保護間隔(Guard Interval)被設置為1/4 ’通道估計時的符號座標轴插值方法使用了線性插值 法。此外，為了頻率軸插值而使用的低通濾波器之截止(Cutoff)頻率為0.2 , 分接頭數為21個。第1G圖為僅存在AWGN _境下進行模擬的結果，而 0 第11圖為在1116通道的都卜勒(Doppler)頻率為10Hz的環境下進行模擬的 結果。 •從第ίο圖以及第u圖可清楚知道，在DVB系統下根據頻率轴插值方 法的不同，其性能是具有差異的。線性插值法雖然具有最為簡便的優點， 但因為幾乎沒有雜崎制效果，因此其雜她之下是最絲。使用低通 濾波器的枝雜比雜減法好，但是也不雜佳崎姆訊雖然增 慮波器之刀接頭數並根據具體狀況來使用最佳的截止鮮時，也能表現 出較好的性能，但每次根據通道賴錢㈣最佳的舰雜數在實際上 非常困難。 在第及第11圖中使用了職方法和未使用增益增強方法^ 上表現出來的性能疋幾乎相闕。使用增益增強方法時，雖然提高了繼 道之__準確度，但在第關以及第關中轉比的制 ',、1/2 ’因此邊緣中一部分的通道估計的錯誤可以通過通道編瑪來杉 從而看不出性能差異。但是在第圖中，根據增益增強方法的使用应 維转出明顯的性能差異。由於在第12圖中使用了 64-Q·及2/3 ^ Z編碼逮率’耻騎估計的通道準確度轉敏感。當未使用增益増 比性估計的通道的絲增益損失雜恢復，因此會發生信號雜訊 月匕间也無法使位元誤差比率下降的情況，性能嚴重的劣化。相反地， 15 201012136 ' 使用增益增強方法時，所估計的通道之邊緣增益損失能很好地被補償，從 而表現出較好的性能。因此增益增強方法是在高速傳送資料的情況下利用 DFT運算來進行通道估計時，為了穩定的信號接收而必須使用的技術。 本發明還可以透過可被電腦讀取的代碼來實現。電腦可讀取的記錄媒 趙包括可被電腦系統讀取的且存有資料的所有種類的記錄裝置。電腦可讀 取的記錄媒體’例如有、RAM、CD ROM、磁帶、軟碟及光資料存 儲裝置等，另外還包括通過載波，例如通過網際網路的傳送狀態來實現。 此外，可被電腦讀取的記錄媒體分散在連接於網路的電腦系統，且以分散 〇 方式存儲並執行電腦可讀取的代碼。 综上所述，雖然本發明已用較佳實施例揭露如上，然其並非用以限定 本發明’本發明所屬技術領域中具有通常知識者，在不脫離本發明之精神 和範圍内’當可作各種之更動與潤飾，因此本發明之保護範圍當視後附之 申請專利範圍所界定者為準。 【圖式簡單說明】 第1圖係繪示DVB系統中’ 1個0FDM符號所包含之導頻之大致形 φ 狀； 第2圖係繪示執行符號座標軸插值時，可知道通道值的位置； 第3圖係鳍示本發明之一實施例之DVB系統； 第4圖係繪示第3圖之!opr運算部之輸出； 第5圖係緣示第3圖之通道脈衝響應估計裝置之功能方塊圖； 第6圖係繪示第4圖之丽運算輸出之放大圖，說明利用臨界值和 預留長度製作剪切視窗的方法； 第7圖係繪示在僅存在有非常小的加法白色高斯雜訊的 DFT運算來對通道進行估計； 〜用 第8圖係繪示將所估計的通道與實際通道兩者之均方誤差值； 201012136 第9圖係_第3圖之增麵強部之邊緣增益補強方法； 〜第1〇圖以及第11圖係繪示透過根據頻率軸插值之維特比(\版加)運 &錯辦進行電雜擬之結果 :以及" Synchronization with data, delaying the symbol, using scattered pilots for time-interpolation is not enough to use only one OFDM symbol. After storing multiple OFDM symbols, use the channel information of the previous symbol. Interpolation is performed, so a delay of the OFDM symbol occurs. That is to say, in order to estimate the channel of the currently received OFDM symbol, an OFDM symbol after a plurality of symbols is also required. The gain enhancement section 318 is for compensating for the gain loss of the estimated channel edge portion. Since the channel impulse response is windowed on the time axis by using the DFT operation, the Sine function is convolved on the frequency axis, causing the gain of the estimated channel edge portion to be attenuated. The edge gain enhancement unit 318 performs inverse compensation on the phenomenon. The demapper 320 performs channel compensation on the signal received after passing through the channel, so that it can be processed to be used in subsequent operations. When the DVB system 300 transmits digital data, it is modulated by one of qPSK, 16-QAM, and 64-QAM as needed. For example, in the simplest qPSK case, 'maps from multiple planes to one point in {1+j, 1-j ′ -1+j. In the DVB system 300, the frequency axis interpolation method is realized by the IDFT unit 3〇6, the channel impulse response estimating unit 308, the window generating unit 310, the second DFT calculating unit 312, and the gain enhancing unit 318. The frequency axis interpolation method performs IDFT operation on the channel information existing in the second picture with the spacing of three subcarriers through symbol coordinate axis interpolation, thereby obtaining the channel impulse response of the 〇FDM symbol through the time axis, but the channel The impulse response has only ^ values in every 3 IDFT inputs, and the others are filled, so as shown in Figure 4, the channel impulse response is repeated three times. Among them, only the i-th rush is filled, and the third pulse of the second side is filled, and then the DKT operation is performed to obtain the result of the frequency axis interpolation. At this time, 2/3 of the total energy in the result of the idft operation is filled with 〇', so it is necessary to multiply the first multiplier 314 by 3 pieces to adjust the entire energy. ° Finding the exact channel impulse response from the results of the IDFT operation will affect the accuracy of the channel being used. In the DVB system, the DFT is used to perform the frequency axis interpolation. 201012136 is considered. The following characteristics: DVB system does not All subcarriers of the size are calculated using the entire Fast Fourier Transform (fine (10) Fine Transform 'hereinafter referred to as FFT). For example, in the case of 2κ's fft operation, Wei__17G5_thief, the rest of the charge, Q, read the idft operation and then send the signal. Since the terminal does not know that the receiving portion is filled with the channel information of G, and therefore, the DFTS calculation can also be used only for the interpolation of the axis. However, this has a side record with the rectangular window (4) plus the analectal window whose length is estimated on the channel value of the fresh axis pair. The shape of the pulse response of the new circuit is not abnormal. It shows the state where the subcarrier spacing is 丨 and the zero crossing point is performed with a period of 2〇48/17()5 (the Zer〇_CroSsing_Sinc function is sampled. It can also be seen from the larger value in Fig. 4) There are repeated and small pulsating components around the pulse. Since these pulsating components are derived from the channel impulse response, these parts should be retained in order to improve the accuracy of the frequency axis interpolation. DFT calculation is used in the system. If only the above characteristics are considered in the frequency axis interpolation, it is better to keep the part that has the largest value from the IDFT operation result as far forward and backward as possible. However, if the health part is too wide, it is equal to the advantage of discarding the bribe operation. In general, the channel pulsing points appear in the narrower part, and the noise components are distributed in the whole position. Therefore, when the interpolation is performed by the DFT operation, the position of the non-channel pulse component is filled. 〇' has the effect of removing the noise component occupying a wide position. The channel lining response tilting method of the present embodiment considers the above-mentioned hiding, # idft operation result and calculates the critical value of the channel pulse component. One of the power and data modulation modes in the D-operation result determines the reserved value, which is used to determine the length of the channel pulse before and after the long-term 戽 is determined by the data modulation method. Please refer to Figure 5 for the silk display. Functional block diagram of the channel impulse response estimating apparatus of the figure: The channel impulse response estimating means 308 includes a power value calculating section 5.2, a power value=average calculating section 504, a second multiplier 5〇6, and a clipping window. The generating unit 5〇8. The power value calculating unit 5〇2 is used to calculate the wheeling power of the Jewish 3〇6 (as shown in Fig. 3). 201012136 Due to the IDFT 3〇6 (as shown in Figure 3) <The output is a complex number, so the power value is calculated and transmitted (a is calculated. The power value average calculation unit 5〇4 is used to obtain an IDFT unit having the same length as the end of the loop word (as shown in Fig. 3) 3 〇6 output power, please do as shown in Figure 3 The average value of the power. The value of the cyclic prefix is determined by the transmitting end. When the DVB system 3〇〇 (such as the 3rd riding) selects 1/4, 1/8, 1/16, and 1/32 of the entire OFDM length. One is transmitted later. The first multiplier unit 506 averages the output of the IDFT unit 306 (as shown in Fig. 3) equal to the length of the cyclic prefix, and the value of the value is __α, which is the towel α. It is determined by the 镰 变 撕 , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , The average value of the power values of the equal length caffeine operation is multiplied by the & value after the threshold value as shown in Fig. 6. The critical value refers to the part of the IDFT operation that finds that the power value exceeds the critical value. The increase reservation means the front and rear parts of the part where the power value exceeds the critical value in the result of the chat operation, and the cut window is set after retaining the range equal to the length of G in Fig. 6. In the present embodiment, after the time synchronization of the FDM symbol is performed for a certain period of time, since the synchronization error of the time 10 is not large, it can be assumed that the channel impulse response component & is included in the cyclic prefix of the 〇FDM symbol. As shown in Fig. 2, when the DFT operation is performed on the channel in the state where the symbol coordinate axis interpolation is completed, the same channel impulse response is repeated three times, and the average value from the front side of the operation result is equal to the length of the first word. Calculation of power. Secondly, different values are obtained according to the modulation method adopted by each data in the OFDM symbol, and the average power is multiplied by the alpha value as a critical value for determining the channel impulse response component. Among them, the unique alpha value obtained by the data modulation method is due to the bit error error Err〇r BER and the signal-to-noise ratio (SNR). Simply put, the signal-to-noise ratio of the signal that can be received when the data modulated by Q SK is transmitted is smaller than the signal-to-noise ratio of the receivable signal of the data modulated by 64-QAM, so the difference is obtained according to the transmission method. Pro 11 201012136 'Boundary value' can get the best performance. The threshold value determined in this way is used as a reference to compare the power of the IDFT operation result and find the position of the channel impulse response component. Then set the Cut Window. The cut window retains the reserved range before and after the position determined as the channel impulse response component, and allows the passage of the range, and does not belong to the range. , to fill. Considering that the channel impulse response is repeated three times in the first! DFT operation result, the second and third components are excluded. As mentioned earlier, the reason for preserving the front and rear portions of the channel impulse response component by the value of the reserved range is to preserve the impulse component based on the impulse response.取得 According to the modulation method of the digital signal, the values of different reserved ranges are obtained for two reasons: the reason that the first value and the alpha value have different values according to the transmission mode are the same; the second is that the QpSK mode and the QAM square are different. In QPSK's sentiment, 卩, 减 叙 她 她 叙 叙 叙 叙 叙 叙 叙 叙 叙 ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' ' The estimated gain loss of the edge portion of the channel should be greater than the reserved value of the QPSK mode. When the IDFT operation result is set as shown in the above window, the channel information inserted by the frequency axis is calculated. When the channel information is interpolated by the above method in the frequency axis, the pulsating component of the channel pulse of the appropriate length according to the modulation method is included, but the length has to be limited in order to remove the noise component. The so-called limit filament setting window is based on the time axis of the channel impulse response on the rectangular window (Rectangular Window) 'DFT operation of the result will result in gain sensitivity (GamLQSS) on the edge of the estimated channel. In order to explain the gain loss more briefly, please refer to Figure 7 for the channel to estimate the channel in the presence of only a very small AWGN $ channel. In Figure 7, set the reserved value when setting the cut window to ι〇. The top graph of Figure 7 does not utilize the real values of the channels estimated by the DFT operation, the middle graph represents the imaginary value, and the bottom graph represents the absolute value of the estimated channel. The channel 12 201012136 - which is estimated when there is no actual passage through the channel, has a imaginary value because the wire will be scaled. It should be noted that the difference between the estimated channel and the edge portion is greater than the actual channel value. For further details, please refer to Fig. 8 for plotting the mean square error value of both the estimated channel and the actual channel. As can be seen from the figure, the estimated channel of the edge portion is very low, and this is due to the setting of the window for the result of the IDFT operation. When the frequency axis interpolation method of the present embodiment is used, in the front part of the __ degree before the channel impulse response portion is judged, the channel lin response component HT is covered by the wire, and the gambling ship is covered. When the ship's face-to-face response is +/-, the edge portion of the channel obtained by DFT is not biased toward real or imaginary numbers, which has only a similar gain loss. If it is not set in the center of the cut window, but the Wei window is set under the bias-side of the lining, the real and imaginary parts of the gain loss value of the edge portion of the _ channel through the DFT operation will be different and will be estimated as It is very difficult to recover the state when the phase is restored. Please refer to Fig. 9 for the edge gain enhancement method of the gain enhancement unit 318 of Fig. 3. Hereinafter, a method of compensating for the gain loss of the size of the triangular oblique line region in the ninth figure of the edge portion of the channel obtained by the DFT operation when the frequency axis is interpolated will be described. ❹ In Figure 9, the value of [b/(a+b)] does not change greatly due to channel conditions or other influences, and remains almost depreciated, but the value of c is the same as the frying window used above. The amplitude is inversely proportional to the variation of the ship's transmission mode and channel conditions. Therefore, each OFDM symbol calculates the c value according to the amplitude of the clipping window, and the edge portion of the estimated channel is Compensation is made in a shape close to a right triangle, thereby improving the performance of the channel estimation. This method only adjusts the size of the estimated channel accurately. Therefore, when using the QAM method sensitive to the channel size, a larger gain can be achieved than using the QpSK method which is relatively insensitive to the channel size. Table 1 shows the mean square error value of the channel edge 2 〇 instantaneous value obtained by performing the frequency axis interpolation by the DFT calculation method of the present invention under various channel conditions and signal noise ratios. GB 〇 ff 13 201012136 in Table 1 indicates that Gain Boosting is not used, while GB On indicates that gain enhancement is used. It can be seen from Table 1 that the accuracy of the channel estimated using the gain enhancement is not related to the channel condition, except for the extremely poor signal noise ratio. Especially when the gain enhancement is not used, even if the signal noise ratio is increased, the accuracy of the estimated channel is hardly increased. The unexpected reduction of the channel filaments in the unsatisfactory condition of the transport table is not related to the hygienic reduction. Conversely, gain enhancement compensates for the gain loss of the edge portion, so it can be confirmed that the larger the signal noise ratio, the better the accuracy of the estimated channel. Table 1 Channel Type SNR(dB) >pc 1 GBOff GBOn AWGN 5 0.12985 0.14930 10 0.05912 0.01222 15 0.05626 0.00575 20 0.05362 0.00384 Portable 5 0.13490 0.15813 10 0.06017 0.02521 15 0.05232 0.01615 20 0.04941 0.01296 SFNSL 5 0.15176 0.19871 10 0.07726 0.04474 15 0.06421 0.02820 20 0.06018 0.02239 SFNSS 5 0.15708 0.16244 10 0.07609 0.03637 15 0.06762 0.02345 201012136 20 0.06547 0.01953 Please refer to the figure ίο and u, showing the Viterbi operation by interpolation based on the frequency axis. The bit error rate is the result of a computer simulation. The basic assumptions of computer simulation are as follows: The system model is based on the DVB system, the FFT size is 8K model, and the guard interval (Guard Interval) is set to 1/4 ′. The symbol coordinate axis interpolation method uses the linear interpolation method. Further, the cutoff (Cutoff) frequency used for the frequency axis interpolation is 0.2, and the number of taps is 21. Fig. 1G shows the results of simulations only in the presence of AWGN_, while Fig. 11 shows the results of simulations performed in an environment where the Doppler frequency of the 1116 channel is 10 Hz. • It is clear from the ίο and u diagrams that the performance differs depending on the frequency axis interpolation method under the DVB system. Although the linear interpolation method has the most simple advantages, it has the most silky effect because it has almost no effect of miscellaneous. The low-pass filter is better than the subtractive method, but it is also not good. Although the number of the knife joints of the wave filter is increased and the best cut-off time is used according to the specific conditions, it can also show better. Performance, but every time according to the channel Lai Qian (four) the best ship clutter is actually very difficult. The performance shown in the first and eleventh figures and the performance of the unused gain enhancement method ^ are almost identical. When the gain enhancement method is used, although the __accuracy of the successor is improved, the system of the first and second turn ratios is ', 1/2', so a part of the channel estimation error in the edge can be programmed through the channel. Cedar does not see performance differences. However, in the figure, the use of the gain enhancement method should shift out significant performance differences. Since the 64-Q· and 2/3 ^ Z coding rate is used in Figure 12, the channel accuracy is estimated to be sensitive. When the gain gain of the channel is not recovered using the gain-ratio estimation, signal noise occurs, and the bit error ratio cannot be lowered between months and the performance is severely deteriorated. Conversely, 15 201012136 ' When using the gain enhancement method, the estimated edge gain loss of the channel is well compensated, resulting in better performance. Therefore, the gain enhancement method is a technique that must be used for stable signal reception when performing channel estimation using DFT calculation in the case of transmitting data at high speed. The invention can also be implemented by means of code that can be read by a computer. The computer readable recording medium includes all kinds of recording devices that can be read by the computer system and stored with data. The computer-readable recording medium 'e.g., RAM, CD ROM, magnetic tape, floppy disk, and optical data storage device, etc., may also be implemented by a carrier wave, for example, via a transmission state of the Internet. In addition, the recording medium readable by the computer is dispersed in a computer system connected to the network, and the computer readable code is stored and executed in a distributed manner. In the above, although the present invention has been disclosed in the above preferred embodiments, it is not intended to limit the invention, and the invention may be made without departing from the spirit and scope of the invention. Various modifications and refinements are made, and the scope of the present invention is defined by the scope of the appended claims. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a diagram showing a general shape of a pilot included in one OFDM symbol in a DVB system; FIG. 2 is a diagram showing a position of a channel value when performing symbol coordinate axis interpolation; 3 is a DVB system showing an embodiment of the present invention; FIG. 4 is a diagram showing the output of the opr operation unit of FIG. 3; and FIG. 5 is a diagram showing the function of the channel impulse response estimating apparatus of FIG. Block diagram; Figure 6 is an enlarged view of the output of the operation of Figure 4, illustrating the method of making a cut window using the critical value and the reserved length; Figure 7 shows that there is only a very small addition white Gaussian noise DFT operation to estimate the channel; ~ Use the 8th figure to show the mean square error value of both the estimated channel and the actual channel; 201012136 Figure 9 _ Figure 3 The edge gain enhancement method; ~1st and 11th drawings show the results of the electrical simulation through the Viterbi (\版加) and the wrong operation according to the frequency axis interpolation:
第12圖係繪示都卜勒頻率為10Hz的TU6通道中，當使用64-QAM 及2/3的維特比編碼速率時，根據頻率轴插值方法之維特比計算後之位元 誤差率圖形。 【主要元件符號說明】 300 dvb系統 302 第一 DFT運算部 304 時間轴插值部 306 IDFT運算部 308 通道脈衝響應估計裝置 310 視窗發生部 312 第二DFT運算部 314 第一乘法器 316 符號延遲部 318 增益增強部 320 解映射器 502 功率計算部 504 功率值平均計算部 506 第二乘法器 508 剪切視窗發生部Fig. 12 is a graph showing the bit error rate after the Viterbi calculation based on the frequency axis interpolation method when the Viterbi coding rate of 64-QAM and 2/3 is used in the TU6 channel with a Doppler frequency of 10 Hz. [Description of main component symbols] 300 dvb system 302 First DFT calculation unit 304 Time axis interpolation unit 306 IDFT calculation unit 308 Channel impulse response estimation device 310 Window generation unit 312 Second DFT operation unit 314 First multiplier 316 Symbol delay unit 318 Gain enhancement unit 320 demapper 502 power calculation unit 504 power value average calculation unit 506 second multiplier 508 cut window generation unit
1717