[go: up one dir, main page]

TWI531171B - Window stop method for turbo decoding - Google Patents

Window stop method for turbo decoding Download PDF

Info

Publication number
TWI531171B
TWI531171B TW102100274A TW102100274A TWI531171B TW I531171 B TWI531171 B TW I531171B TW 102100274 A TW102100274 A TW 102100274A TW 102100274 A TW102100274 A TW 102100274A TW I531171 B TWI531171 B TW I531171B
Authority
TW
Taiwan
Prior art keywords
window
decoding
convergence
core
stop
Prior art date
Application number
TW102100274A
Other languages
Chinese (zh)
Other versions
TW201429170A (en
Inventor
Cheng-Hong Lin
Zhi-Jia Wei
shu-wei Guo
Li-An Ou
Original Assignee
Univ Yuan Ze
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Univ Yuan Ze filed Critical Univ Yuan Ze
Priority to TW102100274A priority Critical patent/TWI531171B/en
Publication of TW201429170A publication Critical patent/TW201429170A/en
Application granted granted Critical
Publication of TWI531171B publication Critical patent/TWI531171B/en

Links

Landscapes

  • Error Detection And Correction (AREA)

Description

用於渦輪解碼之視窗停止方法 Window stop method for turbo decoding

本發明係關於一種用於渦輪解碼之視窗停止方法,尤其是一種用於渦輪解碼演算法,以減少渦輪解碼每一次疊代運算內部視窗之運算量,節省不必要運算進而降低功率消耗。 The present invention relates to a window stop method for turbo decoding, and more particularly to a turbo decoding algorithm for reducing the amount of computation of the internal window of each iteration of the turbo decoding, saving unnecessary operations and reducing power consumption.

現今的通訊系統中,前饋式錯誤更正碼(Forward Error Correction,FEC)在系統中佔了很重要的角色,其主要目的是保護資料。一旦資料在經過錯誤更正碼模組編碼之後,即便在傳送過程中受到傳輸通道裡的雜訊干擾,仍可以在接收端經由解碼的過程後,正確無誤地收到所要傳送的訊息;而渦輪碼(Turbo Code,TC)是現今更正能力較好的前饋式錯誤更正碼之一,渦輪碼是藉由兩個軟式輸入軟式輸出解碼器疊代運算、彼此交換機率資訊,使其編碼增益逼近薛農極限(Shannon Limit),也因其優異之效能,渦輪碼也被廣泛地採用在各無線通訊標準中,例如:DVB、WCDMA、LTE、WiMAX…等,亦可被應用於電力線傳輸系統中,例如:HomePlug。 In today's communication systems, the Forward Error Correction (FEC) plays an important role in the system, and its main purpose is to protect data. Once the data is encoded by the error correction code module, even if it is interfered by the noise in the transmission channel during the transmission process, the message to be transmitted can be received correctly after the decoding process through the decoding process; (Turbo Code, TC) is one of the feedforward error correction codes with better correction ability. The turbo code is the iterative operation of two soft input soft output decoders, and the exchange rate information of each other makes its coding gain close to Xue. Shannon Limit, also because of its excellent performance, turbo code is also widely used in various wireless communication standards, such as DVB, WCDMA, LTE, WiMAX, etc., can also be applied to power line transmission systems. For example: HomePlug.

渦輪碼解碼是藉由軟式輸入軟式輸出(Soft-input Soft-output)演算法解碼器與交錯器彼此之間交換資訊,以進行疊代運算,能夠經過多次疊代(Iterative)運算且相互交換資訊,藉此提高錯誤更正能力。而現行的渦輪解碼器(Turbo Decoder)設計,大多採用視窗化(Window)解碼,以節省大量渦輪解碼延遲並提高渦輪碼吞吐量,然而隨著渦輪解碼器疊代運算次數增加,其消耗之功率也呈線性增加,進而縮短行動裝置使用之時間; 為了解決上述問題,目前已有許多研究提出多種早疊代終結技術(Early Termination Technique),而早疊代終結技術是一套有效偵測多餘疊代次數之技術,能夠提早停止已經收斂之渦輪解碼,並維持渦輪碼整體的解碼效能。然而早疊代終結技術只針對整個資訊框架收斂之狀況下,才停止整個渦輪解碼器之疊代運算,因此在早疊代終結技術條件尚未達成前,渦輪解碼並無法進一步停止內部解碼的運算,且無法偵測內部資訊並停止其運算。 Turbo code decoding is performed by the Soft-input Soft-output algorithm decoder and the interleaver to exchange information with each other for the iterative operation, which can be subjected to multiple Iterative operations and exchanged with each other. Information to improve error correction capabilities. The current turbo decoder (Turbo Decoder) design mostly uses window decoding to save a lot of turbo decoding delay and improve turbo code throughput. However, as the number of turbo decoders is increased, the power consumed is increased. It also increases linearly, which in turn shortens the time it takes to use the mobile device; In order to solve the above problems, many studies have proposed a variety of Early Termination Techniques, and the early iterative termination technique is a technique for effectively detecting the number of superpositions, which can stop the turbo decoding that has already converged. And maintain the overall decoding performance of the turbo code. However, the early iterative termination technique stops the iterative operation of the entire turbo decoder only for the convergence of the entire information framework. Therefore, the turbo decoding cannot further stop the internal decoding operation until the early iteration termination technology condition is reached. It is not possible to detect internal information and stop its operation.

因此,為了減少渦輪解碼每一次疊代運算內部視窗之運算量,以節省不必要運算進而降低功率消耗。在渦輪解碼中,會於軟式輸入軟式輸出解碼後執行視窗偵測,以偵測並記錄解碼資訊內各視窗之收斂狀況,故能於往後之疊代運算,針對已收斂之視窗停止其內部運算進而達到低功耗渦輪解碼之目的,如此應為一最佳解決方案。 Therefore, in order to reduce the amount of computation of the internal window of each iteration of the turbo decoding, unnecessary operations are saved to reduce power consumption. In the turbo decoding, the window detection is performed after the soft input soft output decoding, to detect and record the convergence of the windows in the decoded information, so that the iterative operation can be performed in the future, and the internal window is stopped for the converged window. The operation thus achieves the goal of low power turbo decoding, which should be an optimal solution.

本發明即在於提供一種用於渦輪解碼之視窗停止方法,能夠於渦輪解碼之軟式輸入軟式輸出解碼後執行視窗偵測,以偵測並記錄解碼資訊內各視窗之收斂狀況,故於往後之疊代運算,能夠針對已收斂之視窗停止其內部運算,以減少渦輪解碼每一次疊代運算內部視窗之運算量,節省不必要運算進而降低功率消耗。 The invention provides a window stopping method for turbo decoding, which can perform window detection after decoding the soft input soft output of the turbo decoding, so as to detect and record the convergence state of each window in the decoded information, so in the future The iterative operation can stop its internal operations for the converged window to reduce the amount of computation of the internal window of each iteration of the turbo decoding, saving unnecessary operations and reducing power consumption.

可達成上述發明目的之用於渦輪解碼之視窗停止方法,係於每一次疊代運算中進行渦輪解碼時,當軟式輸入軟式輸出解碼器執行渦輪解碼後,使用視窗偵測器對解碼資訊進行視窗偵測,以偵測並記錄解碼資訊內各視窗之收斂狀況。 The window stop method for turbo decoding which can achieve the above object is a window for decoding information when the soft input soft output decoder performs turbo decoding after performing turbo decoding in each iteration operation. Detect to detect and record the convergence of the windows in the decoded information.

更具體的說,所述軟式輸入軟式輸出解碼器進行軟式輸入軟式輸出解碼後,能夠於視窗內產生解碼位元之軟式或硬式資訊,而該視窗偵測器係將視窗內之解碼位元之軟式或硬式資訊進行比對,以偵測各視窗之收斂狀況。 More specifically, the soft input soft output decoder can generate soft or hard information of the decoding bit in the window after performing soft input soft output decoding, and the window detector is to decode the bit in the window. Soft or hard information is compared to detect the convergence of each window.

更具體的說,所述視窗偵測器判斷整個視窗都達到收斂時,記錄此視窗為收斂之視窗,並於之後的疊代運算中,停止已收斂視窗之運算。 More specifically, when the window detector determines that the entire window has reached convergence, the window is recorded as a window of convergence, and in the subsequent iterative operation, the operation of the converged window is stopped.

更具體的說,所述停止已收斂視窗之運算,能夠使用一可靠資訊做為交換之外部資訊,以提供軟式輸入軟式輸出解碼能夠在停止部分運算下,依舊繼續正常運作並維持其錯誤更正能力。 More specifically, the operation of stopping the converged window can use a reliable information as an exchange of external information to provide a soft input. The soft output decoding can continue to operate normally and maintain its error correction capability under the stop part operation. .

更具體的說,所述停止已收斂視窗之運算,能夠使用一記憶體暫存外部資訊,以提供軟式輸入軟式輸出解碼於停止部分運算下,依舊繼續正常運作並維持其錯誤更正能力。 More specifically, the operation of stopping the converged window can temporarily store external information using a memory to provide soft input soft output decoding under the stop part operation, and continue to operate normally and maintain its error correction capability.

更具體的說,所述解碼資訊中係具有數個平行解碼核心,而該視窗偵測器判斷一個平行解碼核心中所有的視窗皆收斂時,該解碼核心可定義為一個收斂核心,該收斂核心將在之後的疊代運算中停止運算,而一個收斂核心之運算停止等同於停止該收斂核心中所有運算。 More specifically, the decoding information has a plurality of parallel decoding cores, and when the window detector determines that all the windows in a parallel decoding core converge, the decoding core can be defined as a convergence core, and the convergence core The operation will be stopped in the subsequent iterative operation, and the operation of one convergence core stops equivalent to stopping all operations in the convergence core.

更具體的說,所述停止收斂核心之運算,能夠使用一可靠資訊做為交換之外部資訊,以提供軟式輸入軟式輸出解碼能夠在停止部分運算下,依舊繼續正常運作並維持其錯誤更正能力。 More specifically, the operation of stopping the convergence core can use a reliable information as an exchange of external information to provide soft input soft output decoding, which can continue to operate normally and maintain its error correction capability under the stop part operation.

更具體的說,所述停止收斂核心之運算,能夠使用一記憶體暫存外部資訊,以提供軟式輸入軟式輸出解碼能夠在停止部分運算下,依舊繼續正常運作並維持其錯誤更正能力。 More specifically, the operation of stopping the convergence core can use a memory to temporarily store external information to provide soft input. The soft output decoding can continue to operate normally and maintain its error correction capability under the stop part operation.

有關於本發明之前述及其他技術內容、特點與功效,在以下配合參考圖式之較佳實施例的詳細說明中,將可清楚的呈現。 The above and other technical contents, features and advantages of the present invention will be apparent from the following detailed description of the preferred embodiments.

請參閱圖一,為本發明一種用於渦輪解碼之視窗停止方法之主要架構示意圖,由圖中可知,本發明提出視窗式運算停止技術(Window-stopped,WS),為了能有效偵測各位元之收斂情況,於軟式輸入軟式輸出(Soft-input Soft-output,SISO)解碼器1後,插入視窗偵測器2(Window-based Detector,WBD),以偵測單一視窗內各點最大事後機率值是否大於預設之臨界值,最大事後機率值大於某個臨界值時,可以假設其解碼出之資訊是正確的,判斷為收斂;而該視窗偵測器2能夠運行視窗式運算停止技術及核心運算停止技術;而WBD將在之後的疊代運算中,當整個視窗(Window)都達到收斂時,紀錄此視窗為收斂之視窗,並在之後疊代運算中,停止已收斂視窗之運算以減少功耗。 Please refer to FIG. 1 , which is a schematic diagram of a main structure of a window stop method for turbo decoding according to the present invention. As can be seen from the figure, the present invention proposes a window-stopped technique (Window-stopped, WS), in order to effectively detect each element. Convergence, after the Soft-input Soft-output (SISO) decoder 1 is inserted into the Window-based Detector (WBD) to detect the maximum probability of each point in a single window. Whether the value is greater than a preset threshold value, and the maximum event probability value is greater than a certain threshold value, it can be assumed that the decoded information is correct, and the judgment is convergence; and the window detector 2 can run the window operation stop technology and The core operation stops the technique; while the WBD will record the window as a convergence window when the entire window is converged in the subsequent iteration operation, and in the subsequent iterative operation, the operation of the converged window is stopped. Reduce power consumption.

而於渦輪解碼過程中,一前向遞迴運算單元能夠將每個平行解碼框架中的最後一個位元計算並儲存,以等待下一次疊代運算,將其讀出給此次疊代運算中的前向遞迴運算單元當初使值使用;而一後向遞迴運算單元能夠將每個視窗中的最後一個位元計算完並儲存,以等待下一次疊代運算時,讀出上次疊代運算中最後一個位元資料給後向遞迴運算單元當初始值使用;然而,單純的視窗停止運算,並未停止前向遞迴運算單元,其前向運算單元仍需繼續計算,以提供給下一個視窗可靠的初始值;因此,本發明再以視窗式運算停止技術為基礎,延伸出核心運算停止 (Kernel-stopped,KS),若一個MAP解碼核心(Decoding Kernel)中所有的視窗皆收斂,可定義該解碼核心為一個收斂核心,該收斂核心將在之後的疊代運算中停止運算,而一個核心的運算停止如同停止該核心中所有的運算,其中包含視窗中的所有運算以及前向遞迴運算單元;由圖二中可知,若渦輪解碼器中有兩個MAP解碼核心(MAP0、MAP1),每個MAP解碼核心中皆有兩個視窗(Window 0、1及Window 2、3),而MAP0中具有一個收斂視窗(Window 0)及一個未收斂視窗(Window 1),且MAP1中具有兩個收斂視窗(Window 2、Window 3),因此能夠定義MAP1為一個收斂的解碼核心;而於渦輪碼的疊代運算中,如圖三所示,進入本次軟式輸入軟式輸出解碼運算301,根據上次記錄之收斂視窗與收斂核心停止其運算302,而後先使用視窗運算停止技術偵測已收斂視窗303,若判斷有收斂,則記錄視窗收斂狀況304,再使用核心運算停止技術偵測已收斂核心305,若判斷有收斂,則記錄核心收斂狀況306後,再次進入下一次的軟式輸入軟式輸出解碼運算307;若判斷為未收斂之視窗或未收斂之核心,則再次進入軟式輸入軟式輸出解碼運算307;此外,針對停止運算之視窗或核心,由於缺乏用來與下一軟式輸入軟式輸出解碼交換之外部資訊(Extrinsic Information),本發明亦提出利用外部記憶體暫存交換外部資訊或一可靠資訊做為交換之外部資訊,以提供軟式輸入軟式輸出解碼能夠在停止部分運算下,依舊繼續正常運作並維持其錯誤更正能力。 In the turbo decoding process, a forward recursive operation unit can calculate and store the last bit in each parallel decoding frame to wait for the next iteration operation and read it to the iterative operation. The forward recursive computing unit initially uses the value; and a backward recursive computing unit can calculate and store the last bit in each window to wait for the next iteration, read the last stack The last bit data in the generation operation is used as the initial value for the backward recursive operation unit; however, the simple window stops the operation and does not stop the forward recursive operation unit, and the forward operation unit still needs to continue to calculate to provide Give the next window a reliable initial value; therefore, the present invention is based on the windowed operation stop technique, extending the core operation stop (Kernel-stopped, KS), if all the windows in a MAP decoding core (Concoding Kernel) converge, the decoding core can be defined as a convergence core, which will stop the operation in the subsequent iteration operation, and one The core operation stops as if all the operations in the core are stopped, including all the operations in the window and the forward recursive operation unit; as shown in Figure 2, if there are two MAP decoding cores in the turbo decoder (MAP0, MAP1) Each MAP decoding core has two windows (Window 0, 1 and Window 2, 3), while MAP0 has a convergence window (Window 0) and an unconverged window (Window 1), and MAP1 has two Convergence window (Window 2, Window 3), so it can define MAP1 as a convergent decoding core; and in the iterative operation of turbo code, as shown in Figure 3, enter this soft input soft output decoding operation 301, according to The convergence window and the convergence core of the last record stop its operation 302, and then use the window operation stop technique to detect the converged window 303. If it is determined that there is convergence, the window convergence condition 304 is recorded, and then used. The core operation stop technology detects the converged core 305. If the convergence is determined, the core convergence status 306 is recorded, and the next soft input soft output decoding operation 307 is entered again; if it is determined that the window is not converged or the core is not converged, Then enter the soft input soft output decoding operation 307; in addition, for the window or core of the stop operation, the present invention also proposes to utilize the external memory due to the lack of external information (Extrinsic Information) for the soft input decoding exchange with the next soft input. Temporary exchange of external information or a reliable information as an exchange of external information to provide soft input soft output decoding can continue to operate normally and maintain its error correction capability under the stop part of the operation.

而本發明提供WiMAX與HomePlug兩種渦輪碼機制做為實施例,第 一實施例係於WiMAX渦輪碼機制下,採用單一解碼核心,以偵測視窗收斂之狀況,本解碼器採用WiMAX通訊標準之渦輪碼標準,在資訊框架1920位元下,最大疊代次數8次,加性高斯白雜訊通道做為模擬條件,在硬體實現上,本技術採用TSMC 90奈米製程進行合成模擬驗證。在操作頻率250 MHz下,經由Synopsys Design Vision軟體合成及Synopsys Prime Power分析功率消耗後,面積及功率消耗分析如圖四及圖五所示,而圖五中之SISO logic係為軟式輸入軟式輸出邏輯單元,該軟式輸入軟式輸出邏輯單元係包含了前向運算遞迴單元、後向遞迴運算單元及事後機率運算單元;由圖五及圖六可知,在採用本技術後,針對已收斂之視窗,可額外減少約57%之功率消耗,再搭配圖六可知,於Eb/N0 0.4 dB以後,可節省之功率消耗已大於所增加之功率代價。 The present invention provides two turbo code mechanisms of WiMAX and HomePlug as an embodiment. The first embodiment is based on the WiMAX turbo code mechanism and uses a single decoding core to detect the state of window convergence. The decoder adopts the WiMAX communication standard. Turbo code standard, under the 1920-bit information frame, the maximum number of iterations is 8 times, and the additive Gaussian white noise channel is used as the simulation condition. In hardware implementation, this technology uses TSMC 90 nanometer process for synthetic simulation verification. At the operating frequency of 250 MHz, after analyzing the power consumption by Synopsys Design Vision software synthesis and Synopsys Prime Power, the area and power consumption analysis is shown in Figure 4 and Figure 5. The SISO logic in Figure 5 is the soft input soft output logic. Unit, the soft input soft output logic unit includes a forward operation recursive unit, a backward recursive operation unit, and a post-event probability operation unit; as shown in FIG. 5 and FIG. 6, after adopting the technology, for the converged window It can reduce the power consumption by about 57%. It can be seen from Figure 6 that after E b /N 0 0.4 dB, the power consumption saved can be greater than the increased power cost.

而本發明之第二實施例則是於HomePlug渦輪碼機制下,採用多解碼核心,以偵測核心收斂與視窗收斂之狀況,本解碼器採用HomePlug通訊標準之渦輪碼標準,其中使用資訊框架長度128、1088、4160,最大疊代次數8次,加性高斯白雜訊通道做為模擬條件,由圖七中可知,各個資訊框架中所使用的平行解碼核心個數與核心中所包含的視窗數目及視窗長度皆有所不同,由圖中可知,核心中的視窗個數是由視窗長度與平行解碼核心個數所決定的,而視窗長度決定了達成視窗運算停止條件的難易度,核心中的視窗個數也決定了達成核心運算停止條件的難易度。 The second embodiment of the present invention uses a multi-decoding core to detect core convergence and window convergence under the HomePlug turbo code mechanism. The decoder uses the HomePlug communication standard turbo code standard, in which the information frame length is used. 128, 1088, 4160, the maximum number of iterations is 8 times, and the additive Gaussian white noise channel is used as the simulation condition. As shown in Figure 7, the number of parallel decoding cores used in each information frame and the windows included in the core are shown. The number and the length of the window are different. As can be seen from the figure, the number of windows in the core is determined by the length of the window and the number of parallel decoding cores, and the length of the window determines the difficulty of achieving the stop condition of the window operation. The number of windows also determines the difficulty of reaching the core operation stop condition.

本發明為了計算已收斂之視窗與核心所省下的運算量,我們在此定義視窗停止運算率(Window Reduction Rate,WRR)與核心停止運算率(Kernel Reduction Rate,KRR),而總停止率(Total Reduction Rate,TRR)則是WRR 和KRR的總和,由圖八中知,WRR與KRR所停止的視窗並未重複計算,而各信雜比(Eb/N0)下的視窗停止率與核心停止率則如圖中所示;此外,為了分析渦輪解碼器,本發明採用TSMC 90nm製成,Synopsys Design Vision頻率操作在250MHz,而為了省去更多的能源消耗,我們結合視窗式運算停止技術、核心運算停止技術與早疊代終結技術,其中早疊代終結技術偵測已收斂的疊代運算進而停止整個渦輪解碼的運算,而在未收斂的疊代運算中,藉著使用核心運算停止技術偵測已收斂的核心進而停止收斂核心的運算;然而在未收斂的核心中,藉著使用視窗式運算停止技術偵測已收斂的視窗進而停止收斂視窗的運算,其中收斂的視窗、核心並不重疊;請參考圖九,整理了驗證後停止的運算單元實際節省的功率,其中未收斂視窗、收斂視窗與收斂核心皆使用Synopsys PrimePower所量測出的功率,其中包含了其它軟式輸入軟式輸出邏輯單元(SISO logic)、前向遞迴運算單元、後向遞迴運算單元、事後機率運算單元、分支記憶體(BMC)、前向記憶體(SMM)、整個解碼核心(SISO)、整體之功率,而SISO logic係包含有分支運算單元、暫存器與控制訊號,其中其它軟式輸入軟式輸出邏輯單元(SISO logic)並不包含前向運算遞迴單元、後向遞迴運算單元及事後機率運算單元,僅包含剩餘的控制單元,故與圖五的SISO logic(軟式輸入軟式輸出邏輯單元)不同;其最後測試結果由圖九中可知,收斂視窗與未收斂視窗相比約可省64%的功率,而本發明所提出之收斂核心更可節省96.6%的功率。 In order to calculate the amount of computation saved by the converged window and the core, the present invention defines a window reduction rate (WRR) and a core stop reduction rate (KRR), and a total stop rate ( Total Reduction Rate, TRR) is WRR The sum of KRR and KRR is known from Figure 8. The window stopped by WRR and KRR is not repeated, and the window stop rate and core stop rate under each signal-to-noise ratio (Eb/N0) are shown in the figure; In order to analyze the turbo decoder, the invention is made of TSMC 90nm, Synopsys Design Vision frequency is operated at 250MHz, and in order to save more energy consumption, we combine window operation stop technology, core operation stop technology and early iteration termination. Technology, in which the early iterative termination technique detects the converged iterative operation and stops the entire turbo decoding operation, and in the unconverged iterative operation, the core is stopped by using the core computation stop technique to stop the convergence. Core computing; however, in the unconverged core, by using the windowed operation stop technique to detect the converged window and stop the operation of the convergence window, the converged window and core do not overlap; please refer to Figure IX, and sorted out The actual power saved by the computational unit that was stopped after verification, where the unconverged window, the convergence window, and the convergence core were measured using Synopsys PrimePower. Rate, including other soft input soft output logic unit (SISO logic), forward recursive operation unit, backward recursive operation unit, after-the-fact probability unit, branch memory (BMC), forward memory (SMM) The entire decoding core (SISO), the overall power, and the SISO logic system includes the branch operation unit, the register and the control signal, wherein the other soft input soft output logic unit (SISO logic) does not include the forward operation recursive unit The backward recursive arithmetic unit and the after-the-fact probability computing unit only contain the remaining control units, so it is different from the SISO logic (soft input soft output logic unit) of FIG. 5; the final test result is known from FIG. 9 , the convergence window and The unconverged window can save about 64% of the power, and the convergence core proposed by the present invention can save 96.6% of the power.

本發明所提供之一種用於渦輪解碼之視窗停止方法,與其他習用技術 相互比較時,更具備下列優點: A window stop method for turbo decoding provided by the present invention, and other conventional technologies When comparing with each other, it has the following advantages:

1.本發明能夠減少渦輪解碼每一次疊代運算內部視窗之運算量,並節省不必要運算進而降低功率消耗,而在渦輪解碼中,會於軟式輸入軟式輸出解碼後執行視窗偵測,以偵測並記錄解碼資訊內各視窗之收斂狀況,故能於往後之疊代運算,針對已收斂之視窗停止其內部運算進而達到低功耗渦輪解碼之目的。 1. The invention can reduce the calculation amount of the internal window of each iteration operation of the turbo decoding, and save unnecessary calculations to reduce power consumption. In the turbo decoding, the window detection is performed after the soft input soft output decoding, to detect Measure and record the convergence of each window in the decoded information, so it can be used in the subsequent iterations to stop the internal operation of the closed window and achieve low-power turbo decoding.

2.本發明與早疊代終結技術不同,早疊代終止技術能夠停止整個疊代運算,然而早疊代終結技術條件尚未達成前,渦輪解碼無法進一步停止內部解碼的運算;而本發明則是提出視窗式運算停止技術與核心運算停止技術,藉由偵測軟式輸入軟式輸出解碼得到之資訊,進而判斷視窗與收斂核心之收斂狀況,而達到收斂時,則停止視窗與收斂核心之運算以減少功耗,因此能在整個資訊框架收斂前,額外偵測部分收斂之視窗,並停止其運算,以達到低功耗渦輪碼設計。 2. The present invention differs from the early iterative termination technique in that the early iterative termination technique can stop the entire iterative operation. However, before the early iterative termination technique condition is reached, the turbo decoding cannot further stop the internal decoding operation; The window-based operation stop technology and the core operation stop technology are proposed. By detecting the information obtained by the soft input soft output decoding, the convergence state of the window and the convergence core is determined, and when convergence is reached, the operation of the window and the convergence core is stopped to reduce Power consumption, so that the partial convergence window can be detected and the operation can be stopped before the entire information frame converges to achieve low power turbo code design.

3.本發明所偵測已收斂之視窗、已收斂之核心並不相互重疊,故本發明有別於早疊代終結技術,能在整個資訊框架收斂前,額外偵測部分收斂之視窗,並停止其運算,以達到低功耗之設計。 3. The detected convergence window and the converged core of the present invention do not overlap each other, so the present invention is different from the early iteration termination technology, and can additionally detect a partial convergence window before the entire information frame converges, and Stop its operation to achieve a low power design.

藉由以上較佳具體實施例之詳述,係希望能更加清楚描述本發明之特徵與精神,而並非以上述所揭露的較佳具體實施例來對本發明之範疇加以限制。相反地,其目的是希望能涵蓋各種改變及具相等性的安排於本發明所欲申請之專利範圍的範疇內。 The features and spirit of the present invention will be more apparent from the detailed description of the preferred embodiments. On the contrary, the intention is to cover various modifications and equivalents within the scope of the invention as claimed.

1‧‧‧軟式輸入軟式輸出解碼器 1‧‧‧Soft Input Soft Output Decoder

2‧‧‧視窗偵測器 2‧‧‧Window Detector

圖一為本發明用於渦輪解碼之視窗停止方法之主要架構示意圖;圖二為本發明用於渦輪解碼之視窗停止方法之收斂視窗與收斂核心示意圖;圖三為本發明用於渦輪解碼之視窗停止方法之偵測判斷流程圖;圖四為本發明用於渦輪解碼之視窗停止方法之第一實施例之渦輪解碼器面積比較示意圖;圖五為本發明用於渦輪解碼之視窗停止方法之第一實施例之功率消耗比較示意圖;圖六為本發明用於渦輪解碼之視窗停止方法之第一實施例之視窗收斂停止比率示意圖;圖七為本發明用於渦輪解碼之視窗停止方法之第二實施例之資訊框架的平行解碼核心個數與視窗個數示意圖;圖八為本發明用於渦輪解碼之視窗停止方法之第二實施例之信雜比下的WRR與KRR使用符號標準示意圖;以及圖九為本發明用於渦輪解碼之視窗停止方法之第二實施例之未收斂視窗、收斂視窗與收斂核心之功率比較示意圖。 FIG. 1 is a schematic diagram of a main structure of a window stop method for turbo decoding according to the present invention; FIG. 2 is a schematic diagram of a convergence window and a convergence core of a window stop method for turbo decoding according to the present invention; FIG. 3 is a window for turbo decoding according to the present invention; FIG. 4 is a schematic diagram of a turbo decoder area comparison of a first embodiment of a window stop method for turbo decoding according to the present invention; FIG. 5 is a diagram of a window stop method for turbo decoding according to the present invention. FIG. 6 is a schematic diagram of a window convergence stop ratio of a first embodiment of a window stop method for turbo decoding according to the present invention; FIG. 7 is a second diagram of a window stop method for turbo decoding according to the present invention. A schematic diagram of the number of parallel decoding cores and the number of windows in the information frame of the embodiment; FIG. 8 is a schematic diagram of the symbol standard used in the WRR and KRR of the second embodiment of the window stop method for turbo decoding according to the present invention; FIG. 9 is an unconvergent window, convergence window and a second embodiment of a window stop method for turbo decoding according to the present invention; FIG. The core of the comparative diagram of power grabbing.

Claims (7)

一種用於渦輪解碼之視窗停止方法,係於每一次疊代運算中進行渦輪解碼時,當軟式輸入軟式輸出解碼器執行渦輪解碼後,使用視窗偵測器對解碼資訊進行視窗偵測,以偵測並記錄解碼資訊內各視窗之收斂狀況,其中該視窗偵測器判斷整個視窗都達到收斂時,記錄此視窗為收斂之視窗,並於之後的疊代運算中停止已收斂視窗之運算,但於停止已收斂視窗之運算時、並未停止一前向遞迴運算單元,其前向運算單元仍需繼續計算,以提供給下一個視窗可靠的初始值。 A window stop method for turbo decoding is performed when turbo decoding is performed in each iterative operation, and when the soft input soft output decoder performs turbo decoding, the window detector is used to perform window detection on the decoded information to detect Measure and record the convergence of the windows in the decoded information. When the window detector determines that the entire window has reached convergence, the window is recorded as a convergence window, and the operation of the converged window is stopped in the subsequent iterative operation, but When the operation of the converged window is stopped, a forward reciprocating unit is not stopped, and the forward unit still needs to continue to calculate to provide a reliable initial value for the next window. 如申請專利範圍第1項所述用於渦輪解碼之視窗停止方法,其中該軟式輸入軟式輸出解碼器進行軟式輸入軟式輸出解碼後,能夠於視窗內產生解碼位元之軟式或硬式資訊,而該視窗偵測器係將視窗內之解碼位元之軟式或硬式資訊進行比對,以偵測各視窗之收斂狀況。 The window stop method for turbo decoding according to claim 1, wherein the soft input soft output decoder can generate soft or hard information of the decoding bit in the window after performing soft input soft output decoding. The window detector compares the soft or hard information of the decoding bits in the window to detect the convergence of each window. 如申請專利範圍第1項所述用於渦輪解碼之視窗停止方法,其中於並未停止一前向遞迴運算單元的情況下、停止已收斂視窗之運算,能夠使用一可靠資訊做為交換之外部資訊,以提供軟式輸入軟式輸出解碼能夠在停止部分運算下,依舊繼續正常運作並維持其錯誤更正能力。 The window stop method for turbo decoding according to claim 1, wherein the operation of stopping the converged window without stopping a forward reciprocating operation unit can use a reliable information as an exchange. External information to provide soft input soft output decoding can continue to operate normally and maintain its error correction capability under the stop part of the operation. 如申請專利範圍第1項所述用於渦輪解碼之視窗停止方法,其中於並未停止一前向遞迴運算單元的情況下、停止已收斂視窗之運算,能夠使用一記憶體暫存外部資訊,以提供軟式輸入軟式輸出解碼於停止部分運算下,依舊繼續正常運作並維持其錯誤更正能力。 The window stop method for turbo decoding according to claim 1, wherein the memory can be temporarily stored using a memory without stopping a forward reciprocating unit and stopping the operation of the converged window. In order to provide soft input soft output decoding under the stop part of the operation, continue to operate normally and maintain its error correction capability. 如申請專利範圍第2項所述用於渦輪解碼之視窗停止方法,其中該解碼資訊中係具有數個平行解碼核心,而該視窗偵測器判斷一個平行解 碼核心中所有的視窗皆收斂時,該解碼核心可定義為一個收斂核心,該收斂核心將在之後的疊代運算中停止運算,而一個收斂核心之運算停止等同於停止該收斂核心中所有運算。 The window stop method for turbo decoding according to claim 2, wherein the decoding information has a plurality of parallel decoding cores, and the window detector determines a parallel solution. When all the windows in the code core converge, the decoding core can be defined as a convergence core, which will stop the operation in the subsequent iterative operation, and the operation of a convergence core stops is equivalent to stopping all operations in the convergence core. . 如申請專利範圍第5項所述用於渦輪解碼之視窗停止方法,其中停止收斂核心之運算,能夠使用一可靠資訊做為交換之外部資訊,以提供軟式輸入軟式輸出解碼能夠在停止部分運算下,依舊繼續正常運作並維持其錯誤更正能力。 The window stop method for turbo decoding according to claim 5, wherein the operation of stopping the convergence core can use a reliable information as an exchange of external information to provide a soft input soft output decoding capable of stopping part of the operation. , continue to operate normally and maintain its ability to correct errors. 如申請專利範圍第5項所述用於渦輪解碼之視窗停止方法,其中停止收斂核心之運算,能夠使用一記憶體暫存外部資訊,以提供軟式輸入軟式輸出解碼能夠在停止部分運算下,依舊繼續正常運作並維持其錯誤更正能力。 The window stop method for turbo decoding according to claim 5, wherein the operation of stopping the convergence core can use a memory to temporarily store external information to provide soft input. The soft output decoding can still be performed under the stop part operation. Continue to function properly and maintain its ability to correct errors.
TW102100274A 2013-01-04 2013-01-04 Window stop method for turbo decoding TWI531171B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
TW102100274A TWI531171B (en) 2013-01-04 2013-01-04 Window stop method for turbo decoding

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
TW102100274A TWI531171B (en) 2013-01-04 2013-01-04 Window stop method for turbo decoding

Publications (2)

Publication Number Publication Date
TW201429170A TW201429170A (en) 2014-07-16
TWI531171B true TWI531171B (en) 2016-04-21

Family

ID=51726235

Family Applications (1)

Application Number Title Priority Date Filing Date
TW102100274A TWI531171B (en) 2013-01-04 2013-01-04 Window stop method for turbo decoding

Country Status (1)

Country Link
TW (1) TWI531171B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TWI672911B (en) * 2019-03-06 2019-09-21 瑞昱半導體股份有限公司 Decoding method and associated circuit

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TWI672911B (en) * 2019-03-06 2019-09-21 瑞昱半導體股份有限公司 Decoding method and associated circuit

Also Published As

Publication number Publication date
TW201429170A (en) 2014-07-16

Similar Documents

Publication Publication Date Title
US7454684B2 (en) Apparatus and method for turbo decoder termination
CN104579369B (en) A kind of Turbo iterative decodings method and code translator
CN101951266B (en) Turbo parallel decoding method and decoder
CN102111162B (en) Turbo component decoding method, component decoder, branch calculator and Turbo decoder
CN102571108B (en) Self-adaptive iterative decoding method for Turbo product codes
CN102523076A (en) Universal and configurable high-speed Turbo code decoding system and method thereof
CN104767537B (en) A kind of Turbo interpretation methods for OFDM electric line communication systems
CN103124181A (en) Turbo code decoding iteration cease method based on cosine similarity
CN101674094A (en) Turbo code channel adaptive implementation method
CN101106380A (en) A method and device for iterative decoding of LDPC codes
TWI531171B (en) Window stop method for turbo decoding
US8918695B2 (en) Methods and apparatus for early stop algorithm of turbo decoding
CN111224676B (en) Self-adaptive serial offset list polarization code decoding method and system
CN102832954B (en) Turbo code iterative decoding stopping method based on soft information average minimum value
US10461776B2 (en) Device and method of controlling an iterative decoder
US8984377B2 (en) Stopping methods for iterative signal processing
TWI504162B (en) A layer operation stop method for low density parity check decoding
CN112398486A (en) Turbo code interleaving parameter identification method and system optimized by using multiple error correction methods
Huang et al. Information theoretic criterion for stopping turbo iteration
CN105515591B (en) A kind of Turbo code decoding system and method
TWI555339B (en) Overlay decoding device, iterative signal testing device and information updating method
CN100486235C (en) Iterative receiving method for maintaining soft information
CN103701475B (en) Decoding method for Turbo codes with word length of eight bits in mobile communication system
CN106330206A (en) A decoding method and device
Mohamad et al. Early stopping turbo iteration at low SNR for CE-based stopping criteria

Legal Events

Date Code Title Description
MM4A Annulment or lapse of patent due to non-payment of fees