TW320794B - Signal processing method and device of interpolation digital filter - Google Patents

Signal processing method and device of interpolation digital filter Download PDF

Info

Publication number
TW320794B
TW320794B TW83108068A TW83108068A TW320794B TW 320794 B TW320794 B TW 320794B TW 83108068 A TW83108068 A TW 83108068A TW 83108068 A TW83108068 A TW 83108068A TW 320794 B TW320794 B TW 320794B
Authority
TW
Taiwan
Prior art keywords
filter
value
aforementioned
output
circuit
Prior art date
Application number
TW83108068A
Other languages
Chinese (zh)
Inventor
Yeong-Jou Perng
Original Assignee
Ind Tech Res Inst
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 Ind Tech Res Inst filed Critical Ind Tech Res Inst
Priority to TW83108068A priority Critical patent/TW320794B/en
Application granted granted Critical
Publication of TW320794B publication Critical patent/TW320794B/en

Links

Abstract

A digital filter, a L phase interpolation, generates the input sampling x([(t-l)/L]-i) according to the n coefficients of tap Pl(i) of L banks in which 1<=l<=L, 0<=i<=n-1. It includes 1. the first circuit to generate bankfilter terms at each time t which includes at least one combination of the two cross symmetric terms Under in-phase condition, Pk(i)x([(t-k)/L]-i)+PL-k+1(i).x([(t-(L-k+1))/L]-i) and Pk(n-1-i)x([(t-k)/L]-(n-1-i))+PL-k+1(n-1-i).x([(t-(L-k+1))/L]-(n-1-i)) Under non in-phase condition, Pk(i)x([(t-k)/L]-i)+PL-k+1(i-1)x([(t-(L-k+1))/L]-(i-1)) and Pk(n-i)x([(t-k)/L]-(n-i))+PL-k+1(n-1-i)x([(t-(L-k+1))/L]-(n-1-i)) 2. the second circuit to sum the above bankfilter terms at each time t for generating the output of the variation from time t-1 to t.

Description

320794 A? B7 五、發明説明(I ) 本發明領域 本項發明係關於使用一個有限脈衝反應濾波器( finite impulse response (FIR) filter)來處理數位訊 號。 本發明的背景 有限脈衝反應濾波器是由若干個分接點(tap)係數或 加重器(weights)組成的數位(digital)濾波器。樣本 中的一個輸入訊號(input signal) x(t)被移入(' shifted into)有限脈衝反應濾波器,每一個循環(per cycle) —個取樣。在每一個循環t,此有限脈衝反應濾 波器(the FIR filter,以下簡稱FIR濾波器)計算一 個和(sum) y(t): n-1 Y(t) = Σ p(i) x (t-i) (1) (請先閣讀背面之注意事項再填寫本頁) ---^----τ--.---f —装--------訂----- 經濟部中央標準局員工消費合作社印裝 th 式(1)中,x(t-i)是一個X的第t-i個取樣(t-i sample of x),p(i)是一個FIR濾波器的第i個分接點 係數,其中〇 S i S (n-1),且n是FIR濾波器内分接點 係數的總量。 FIR濾波器在若干應用上非常有用。例如,增頻化( interpolation)。第一圖所示係一個多相位增頻器1〇〇 , 曾被J. Candy與G. Temes發表在I.E.E.E第 69卷第 3期,I988年3月第417至448頁中的&quot;超取樣資料 線 本紙張尺度適用中國國家揉準(CNS ) A4規格(210X297公釐) 83. 3. !〇,〇〇〇 32〇794 A7 B7 五、發明説明(2) 轉換器:理論,設計及模擬,Oversampling Delta-Sigma Data Converters : Theory,Design and Simulation'· 〇 此增頻器100事實上是由L排(bank) FIR濾波器 110-1,110-2,. · .,110-L所組成且此增頻器輸入訊號 t320794 A? B7 V. Description of the invention (I) Field of the invention This invention relates to the use of a finite impulse response (FIR) filter to process digital signals. Background of the invention A finite impulse response filter is a digital filter composed of several tap coefficients or weights. An input signal x (t) in the sample is shifted into a finite impulse response filter, one sample per cycle. At each cycle t, the finite impulse response filter (the FIR filter, hereinafter referred to as FIR filter) calculates a sum (sum) y (t): n-1 Y (t) = Σ p (i) x (ti ) (1) (Please read the precautions on the back first and then fill in this page) --- ^ ---- τ --.--- f —install -------- order ----- In the th (1) printing form of the employee consumer cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs, x (ti) is the ti sample of x (ti sample of x), and p (i) is the i-th branch of a FIR filter Contact coefficient, where σ S (n-1), and n is the total number of contact coefficients in the FIR filter. FIR filters are very useful in several applications. For example, interpolation. The first picture shows a multi-phase multiplier 100. It was published by J. Candy and G. Temes in IEEE Volume 69, Issue 3, March I988, pages 417 to 448 of "oversampling." The data paper format is applicable to China National Standard (CNS) A4 (210X297mm) 83. 3. 〇〇〇〇〇〇〇 32〇794 A7 B7 V. Description of the invention (2) Converter: theory, design and simulation , Oversampling Delta-Sigma Data Converters: Theory, Design and Simulation '. 〇 This frequency multiplier 100 is actually composed of L bank FIR filters 110-1, 110-2,..., 110-L And the input signal t of this multiplier

XX

C )是經由L倍的增頻信號處理,(其中C) is processed by an L-fold frequency-increasing signal, (where

L 表示最接近且小於等於C的整數)。 第二圖,(a)至(e)在説明取樣値X( t tL represents the nearest integer less than or equal to C). In the second figure, (a) to (e) illustrate the sampling value X (t t

L 的 移入情形,其中取樣値X( )是經L倍超取樣的The shifting situation of L, where the sampling value X () is oversampled by L times

L 經濟部中央標準局®C工消費合作社印製 取樣/保持資料。假設L=5且n=3。定義一個値V=t模 數(mod) L。當時間t=0,v=0時,取樣値見第二(a)圖。 如圖所示,每一個濾波器排從110-1至110-5的値都是 相同的。此時,我們可以説濾波器排110-1至110-5是 處在同相位。當時間t=l,(v=l)如第二(b)圖所示,取 樣値χ(2)將被位移入濾波器排110-1中。因此,濾波 I----7--:--1 I裝-----^-I訂-----ί 線 (請先閲讀背面之注意事項再填寫本頁) 本紙張尺度逋用中國國家揉準(CNS ) Α4規格(210Χ297公釐) 83. 3.10,000 A7 B7 五、發明説明(3 ) 器排110-1與110-2至110-5處於非同相位(out of phase)。相似地,第二(c)圖爲在時間t=2,對v=2時 x(2)位移入濾波器排110-2的情形。第二(d)圖是在時 間t=3對v=3時x(2)位移入濾波器排110-3的情形。 第二(e)圖係在時間t=4,對v=4時,x(2)位移入濾波器 排110-4的情形。最後,第二(f)圖表示在時間t=5,對 v=0時,χ(2)位移入濾波器110-5的情形。在2(f)圖中。 .瀘波器排110-1至110-5將再一次的處於同相位。 每一個 FIR 濾波器排 110-1,110-2,..·.,110-L的 產出値(output)被輸入(fed to)至一個相對應的提高 取樣(upsampling)電路 120-1,120-2,..,120-L,此 電路是將輸出訊號提高取樣L倍。見D. ELLIOT,數位訊 號處理應用工程手册(HANDBOOK OF DIGITAL SIGNAL PROCESSING .ENGINEERING APPICATIONS)第 239 頁(1987) 〇 這些被提高取樣及濾波的訊號yi (t),y2 (t),..., YL (t)將被相加在一起,以便在時間t時,產生輸入訊 號的增頻訊號g(t)。在此每一個提高取樣濾波訊號,例 如,yL (t),是延遲相對應於前一個提高取樣濾波訊號, 例如,yl - jl (t),一個循環單位z-1的產出値。其中, 增頻訊號g(t)可由式(2)表示: ί-------- ^------ΐτ------一紙 (請先閲讀背面之注意事項再填寫本頁) 經濟部中央標準局®c工消費合作社印製 L L n-1 g(t)= Eye(*-i+〇= Σ Σ P (i) 4=1 1=1 i=o Λ t-1L Central Bureau of Standards ® Ministry of Economic Affairs ® C Industrial and Consumer Cooperatives Print sample / hold data. Suppose L = 5 and n = 3. Define a value V = t modulus (mod) L. When time t = 0, v = 0, the sampling value is shown in the second (a) graph. As shown in the figure, the value of each filter bank from 110-1 to 110-5 is the same. At this time, we can say that the filter rows 110-1 to 110-5 are in the same phase. When time t = l, (v = l) as shown in the second (b) graph, the sampling value χ (2) will be shifted into the filter bank 110-1. Therefore, filter I ---- 7-:-1 I installed ----- ^-I set ----- ί line (please read the precautions on the back first and then fill out this page) Using the Chinese National Standard (CNS) Α4 specification (210Χ297 mm) 83. 3.10,000 A7 B7 5. Description of the invention (3) The rows 110-1 and 110-2 to 110-5 are out of phase ). Similarly, the second (c) graph shows the situation where x (2) is shifted into the filter bank 110-2 at time t = 2 and v = 2. The second (d) diagram is the case where x (2) is shifted into the filter bank 110-3 at time t = 3 versus v = 3. The second (e) graph is the case where x (2) is shifted into the filter bank 110-4 at time t = 4 and v = 4. Finally, the second (f) graph shows the situation where χ (2) is shifted into the filter 110-5 at time t = 5 and v = 0. In Figure 2 (f). . The wave filter rows 110-1 to 110-5 will again be in the same phase. The output values of each FIR filter bank 110-1, 110-2, ..., 110-L are fed to a corresponding upsampling circuit 120-1, 120-2,..., 120-L, this circuit is to increase the output signal by sampling L times. See D. ELLIOT, Handbook of Digital Signal Processing Application Engineering (HANDBOOK OF DIGITAL SIGNAL PROCESSING.ENGINEERING APPICATIONS) page 239 (1987). These signals with increased sampling and filtering are yi (t), y2 (t), ..., YL (t) will be added together to generate the frequency-increasing signal g (t) of the input signal at time t. Here, each up-sampling filter signal, for example, yL (t), is the delay corresponding to the previous up-sampling filter signal, for example, yl-jl (t), the output value of a cycle unit z-1. Among them, the frequency-increasing signal g (t) can be expressed by formula (2): ί -------- ^ ------ lτ ------ a piece of paper (please read the precautions on the back first (Fill in this page) LL n-1 g (t) = Eye (*-i + 〇 = Σ Σ P (i) 4 = 1 1 = 1 i = o Λ t-printed by the Central Bureau of Standards of the Ministry of Economic Affairs and Industry and Consumer Cooperatives 1

L -i) (2) 但問題是增頻器100需要許多FIR滹波器,因而在 一個積體電路晶片中將佔據寶貴的空問。第三圖表示另外 83.3. 10,000 本紙張尺度適用中國國家標準(CNS ) Α4規格(210Χ297公釐) 320794 A7 B7 五、發明説明(斗) 一種傳統應用於聲訊(audio)的多相位增頻器2〇〇。在 此一增頻器200中,是假設: t' x( .i) -0L -i) (2) However, the problem is that the frequency multiplier 100 requires many FIR filters, so it will occupy valuable space in an integrated circuit chip. The third picture shows another 83.3. 10,000 The paper scale is applicable to the Chinese National Standard (CNS) Α4 specification (210Χ297 mm) 320794 A7 B7 5. Description of the invention (bucket) A multi-phase frequency multiplier traditionally used in audio 2 〇〇. In this frequency multiplier 200, it is assumed that: t 'x (.i) -0

L 對S关P ,其中,若(t模數L)关0,s = t模數L 且,若t模數L=0,S=L。假如輸入訊號X是取樣資料 上述假設是成立的。因此,方程式(2)可被筒化爲: g(t)= ys (t) n-1 .=0 (i) · x( (請先聞讀背面之注意事項再填寫本頁) t-s •i) (4) 經濟部中央標準局貝工消费合作社印製 在第三圖,濾波器200是伴隨著一個隨機存取記憶體 (RAM) 210,爲了要儲存η個輸入取樣値,例如,χ(〇), x(l),...,x(n-l)。每一個取樣假設爲b位元數。因此, 此 RAM 210 是一個 nX b 位元 RAM。一個(LX η) X d 位 元唯讀記憶體(ROM) 220是用來儲存每一個可能d位元 分接點係數Ps (i)(其中,11SSL且〇SiSn-l)。爲計算 在一個特定時間t的g(t)値,一個控制器(没被列出) 連續地輸出η個適當的取樣値x( t-sL versus S is related to P, where if (t modulus L) is off 0, s = t modulus L and if t modulus L = 0, S = L. If the input signal X is sampled data, the above assumption is true. Therefore, equation (2) can be reduced to: g (t) = ys (t) n-1. = 0 (i) · x ((please read the precautions on the back before filling this page) ts • i ) (4) Printed in the third figure by the Beigong Consumer Cooperative of the Central Bureau of Standards of the Ministry of Economy, the filter 200 is accompanied by a random access memory (RAM) 210, in order to store n input sample values, for example, χ ( ○), x (l), ..., x (nl). Each sample is assumed to be b bits. Therefore, this RAM 210 is an nX b-bit RAM. A (LX η) X d bit read-only memory (ROM) 220 is used to store each possible d bit tap point coefficient Ps (i) (where, 11SSL and 〇SiSn-1). To calculate the g (t) value at a specific time t, a controller (not listed) continuously outputs η appropriate sampling values x (t-s

L i)及η 個相對應的分接點係數PS (i)至相乘器230。此相乘器 23〇將每個取樣値與其相對應的分接點係數相乘且輸出 本紙張尺度適用中國國家標準(CNS ) A4規格(210X297公釐) 83. 3.10,000 妓〇7q 4 A7 B7 五、發明説明(5&quot;) 每一個乘積,至一個累計器24〇。此累計器24〇累計η 個如此的乘積,以產生每一個g(t)。增頻器200因此使 用η個相乘及η個累計來產生g(t)。 若輸入訊號x爲取樣&quot;呆持(sample/hold)資料, ;則方程式(2)可被簡化如下: t k n-2 t g(t)= χ( --- )· Σ P{ (〇)+ Σ X( — L L」 Si = 1 i=0 L L - t L Pi (i+1)] + x( -η) · Σ Pi ( L k •i -1) · [ Σ Pi (i) + Σ p =K+1 q =1 L =k+i (5) _HJ« ^^^1 ^i_l tm I ^^^1 ^^1 (請先聞讀背面之注意事項再填寫本頁)L i) and n corresponding tap coefficients PS (i) to the multiplier 230. This multiplier 23〇 multiplies each sampling value with its corresponding tap point coefficient and outputs the paper size applicable to the Chinese National Standard (CNS) A4 specification (210X297 mm) 83. 3.10,000 prostitute 〇7q 4 A7 B7 5. Description of invention (5 &quot;) Each product goes to a totalizer 24. This accumulator 240 accumulates n such products to produce each g (t). The frequency multiplier 200 therefore uses n multiplications and n accumulations to generate g (t). If the input signal x is sample &quot; sample / hold data, then equation (2) can be simplified as follows: tk n-2 tg (t) = χ (---) · Σ P {(〇) + Σ X (— LL ″ Si = 1 i = 0 LL-t L Pi (i + 1)] + x (-η) · Σ Pi (L k • i -1) · [Σ Pi (i) + Σ p = K + 1 q = 1 L = k + i (5) _HJ «^^^ 1 ^ i_l tm I ^^^ 1 ^^ 1 (please read the precautions on the back before filling this page)

L 其中,HK5_L,且 Σ Pjj (i)=〇,若K=L。此傳統增頻 J? =κ+1 器200可被修改,來執行如此一個取樣/保持增頻器。 特別的是,一個(n+l)Xb位元RAM 210被用來儲存n+1 個輸入取樣値x( t ),x( t —L」 L L」 1)L where, HK5_L, and Σ Pjj (i) = 〇, if K = L. The conventional frequency multiplier J? = Κ + 1 can be modified to implement such a sample / hold frequency multiplier. In particular, a (n + 1) Xb bit RAM 210 is used to store n + 1 input sample values x (t), x (t —L ″ L L ″ 1)

1T 經濟部中央標準局員工消費合作社印製 t x(1T Printed by the employee consumer cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs t x (

LL

-n)。相同地,一個(LX (n+1)) Xd 位元 ROM 220也被用來儲存LX (n+1)個不同的分接點係數和,在 此方程式(5)中,三個分接點係數和計算方程式分別爲 本紙張尺度適用中國國家標準(CNS ) Α4規格(210Χ297公釐) 83. 3.10,000 五、發明説明(6 ) A7 B7 (5a),(Sb),及(sc),其所需相對應儲存記憶體大小分 別爲 L,(n-1) X L 及 L。 _ k Σ Ρ$ (Ο) (5a) L k Σ Pi (i) + Σ } =K+1 A =-n). Similarly, a (LX (n + 1)) Xd bit ROM 220 is also used to store the sum of LX (n + 1) different tap coefficients. In this equation (5), three tap points Coefficients and calculation equations are applicable to the Chinese National Standard (CNS) Α4 specification (210Χ297 mm) for this paper scale 83. 3.10,000 V. Description of the invention (6) A7 B7 (5a), (Sb), and (sc), The required storage memory sizes are L, (n-1) XL and L, respectively. _ k Σ Ρ $ (Ο) (5a) L k Σ Pi (i) + Σ) = K + 1 A =

Pi (i+1) (5b) L Σ Pj? (n-1) » =k+l (5c) 經濟部中央標準局貝工消費合作社印裝 爲了要§1*算每一時間t的每一個g(t)値, (没被顯示)連續地自RAM έ 10輸出n+l個取樣値以及 自ROM 22〇輸出n+1個相對應的分接點係數和至乘法 器23〇。拫據方程式(5),爲了要產生g(t),累計器24〇 要累計n+1個乘法器23〇的乘積。在此情況下,增頻 器200必須執行n+1次的相乘及n+l次的累計。 第四圖係另外一種用來增頻聲訊(audio)取樣資料 的傳統增頻器300。如同前述,增頻器300有一個nXb 位元 RAM 310 和一個(LX η) X b 位元 ROM 3 2.0。不 同於前述的是,每一個儲於ROM 32〇的分接點係數Rjj (i) 被選擇爲有著僅二個或三個的非零(non-zero)位元的二 次幕(a power of two)。在實際執行中,η個取樣値是連 個控制器 83.3· 10,000 I---·------^ —裝-----1--訂-----^ 線 (請先閲讀背面之注意事項再填寫本頁) A7 B7 五、發明説明(7) 續地被自RAM 31〇輸出至一個位移暫存器(shift register) 330 〇 此相對應的η個分接點係數被連接地輸出至一個控 制電路340。根據Booth的運算法則(A1g〇rithm),此 控制電路將每一個分接點係數與其相對應的取樣値相乘。 亦即,對每一個非零位元q,其中〇5_qSd-1 ,取樣値被向 左位移q個位元,且使用累計器3 5〇累計每—個被位移 的輸出結果。此累計器35〇是在一個特定時間t ,累計 η個乘稹以產出g(t)。 第五圖係表示另外一個被揭露(disclosed)於美專 利號碼4,862,4〇2中的傳統增頻器400。此增頻器400 ----.—----ί -裝— (請先閲讀背面之注$項再填寫本頁) 是增頻視訊(Video)取樣資料。此一取樣資料X -i)可被寫成二進位表示法如下: X( t-s -i) b-l =Σ Χχη ( t-s L L」 m=0 L L」 •i) · 2 m t-sPi (i + 1) (5b) L Σ Pj? (N-1) »= k + l (5c) Printed by the Beigong Consumer Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs in order to §1 * count every time t The g (t) value, (not shown) continuously outputs n + 1 sampling values from the RAM 10 and n + 1 corresponding tap coefficient sums from the ROM 22〇 to the multiplier 23〇. According to equation (5), in order to generate g (t), the accumulator 24〇 accumulates the product of n + 1 multipliers 23〇. In this case, the frequency multiplier 200 must perform n + 1 times of multiplication and n + 1 times of accumulation. The fourth diagram is another conventional frequency multiplier 300 for upsampling audio sampling data. As before, the frequency multiplier 300 has an nXb bit RAM 310 and a (LX η) X b bit ROM 3 2.0. Unlike the foregoing, each tap coefficient Rjj (i) stored in the ROM 32 is selected as a secondary screen with only two or three non-zero bits (a power of two). In actual implementation, η sampling values are connected to a controller 83.3 · 10,000 I --- · ------ ^ —installed ----- 1--order ----- ^ line (please first Read the precautions on the back and fill in this page) A7 B7 5. Description of the invention (7) It is continuously output from RAM 31〇 to a shift register (shift register) 330. The corresponding n tap coefficients are The ground is connected to a control circuit 340. According to Booth's algorithm (Algorithm), this control circuit multiplies each tap point coefficient with its corresponding sampling value. That is, for each non-zero bit q, of which is 5_qSd-1, the sampling value is shifted to the left by q bits, and each shifted output result is accumulated using an accumulator 35. This accumulator 35〇 accumulates η multiplying rice at a specific time t to produce g (t). The fifth diagram shows another conventional frequency multiplier 400 that was disclosed in US Patent No. 4,862,402. This upconverter 400 ----.—---- ί -installed — (please read the $ item on the back and then fill in this page) is the sampling data of up-converted video (Video). This sampled data X -i) can be written as a binary representation as follows: X (t-s -i) b-l = Σ Χχη (t-s L L ″ m = 0 L L ″ • i) · 2 m t-s

L (6)L (6)

iT 經濟部中央標準局員工消費合作社印策 其中,Xm是一個代表輸入取樣値爲位元,0 '或,i,的丨 量(Vector),將方程式(4)及(6)結合,g(t)可被寫】 如下: b-l g(t)= Σ m=0 n-l Σ ps i=〇 (i) * Xm ( t-s i)iT The Ministry of Economic Affairs Central Standards Bureau Staff Consumer Cooperative printed it, where Xm is a vector representing the input sampling value as a bit, 0 'or, i, combining equations (4) and (6), g ( t) can be written] as follows: bl g (t) = Σ m = 0 nl Σ ps i = 〇 (i) * Xm (ts i)

L 83. 3.10,000 本紙張尺度適用中國國家榇準(CNS ) A4規格(210X297公釐) 五、發明説明(8 ) A7 B7 方程式(7)被用來執行使用Booth運算法則,計算每—取 樣與他相對應的分接點係數乘積是非常有用的。(此運算法 則是依分接點係數的第q位元,決定每一取樣訊號位移, 然後累計此已位移之取樣訊。然而,增頻器400使用修改 Booth運算法則。在此一修改後的Booth運算法則中,分 接點係數内相鄰的第qi ,q2位元對,〇Sqi 同時被 用來控制累計此向左位移了 qx及q2 =qi +1位元的取樣 訊號。將輸入訊號X所需的從m=0至b-Ι加法運算,依 相鄰二位元平面加以分離,並結_合成平行的偶數或奇數個 加法運算。此時方程式(7)可被重寫如下: b 2 2 · r+1 n-1 g(t)= Σ [ Σ [ Σ Ps (i) · xm r=0 τφ2 · r i=0 t-s i) · 2m)] (8)L 83. 3.10,000 This paper scale is applicable to the Chinese National Standard (CNS) A4 (210X297 mm). V. Description of the invention (8) A7 B7 Equation (7) is used to perform the use of Booth algorithm to calculate each sample The product of tap coefficients corresponding to him is very useful. (This algorithm is based on the qth bit of the tap coefficient to determine the displacement of each sampled signal, and then accumulates the shifted sampled signal. However, the frequency booster 400 uses a modified Booth algorithm. In this modified In the Booth algorithm, the adjacent qi, q2 bit pairs within the tap coefficients, 〇Sqi are also used to control the accumulation of the sampling signal shifted to the left by qx and q2 = qi + 1 bit. The input signal The addition operation from m = 0 to b-1 required by X is separated according to adjacent two-bit planes and combined into parallel even or odd addition operations. At this time, equation (7) can be rewritten as follows: b 2 2 · r + 1 n-1 g (t) = Σ [Σ [Σ Ps (i) · xm r = 0 τφ2 · ri = 0 ts i) · 2m)] (8)

L 在增頻器4〇〇中,一個模組410被提供用來先前處 理(pre-preprocessing)分接點係數ps (i)及將處理 過的結果輸出至模組420-0,420-1,...,420-R,其中 b | ^ γ I裝 訂X線 (請先閲讀背面之注意事項再填寫本頁) R: •1°經由線411每一個分接點係數PS (i)被連 經濟部中央標準局貝工消费合作社印製 續地載入(loaded into)模組410。此η個分接點係數 Ps (i)被位移入第一個位移暫存器的第412欄。進一步 地,乘法器414和加法器416對每一個係數計算 3 · ps (i),且連接地將此n個3 . Ps (i)値位移入位 本紙張尺度適用中國國家榇準(CNS ) A4規格(210X297公釐) 83. 3.10,000 A7 B7 五、發明説明(?) 移暫存器的第418欄一旦完全載入,依特定的S値,模組 410將平行地輸出每一個從i=〇至n-l的成對係數 Ps (i),3 · Ps (i)。更進一步的,依每個S値,位移暫 存器的412欄及418欄是用來提供L· η個暫存器(例 如,412-0-1,.··,412-0-L)以便分別儲存係數ps (i), 3 · PS (i)。另外,多工器(例如,413-1,及415-1)則 輸出適當的係數對,而此係數對(視輸入取樣値 t-s X( •i)而定。L In the frequency multiplier 400, a module 410 is provided for pre-preprocessing tap coefficient ps (i) and outputs the processed result to modules 420-0, 420-1 , ..., 420-R, where b | ^ γ I bind X-rays (please read the precautions on the back before filling in this page) R: • 1 ° each tap point coefficient PS (i) via line 411 is The Beigong Consumer Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs printed and loaded into the module 410 continuously. The n tap coefficients Ps (i) are shifted into column 412 of the first shift register. Further, the multiplier 414 and the adder 416 calculate 3 · ps (i) for each coefficient, and connect the n 3. Ps (i) values into place. The paper scale is applicable to the Chinese National Standard (CNS) A4 specification (210X297mm) 83. 3.10,000 A7 B7 V. Description of invention (?) Once column 418 of the shift register is fully loaded, the module 410 will output each slave in parallel according to the specific S value i = 〇 to nl pair coefficient Ps (i), 3 · Ps (i). Furthermore, according to each S value, the 412 and 418 columns of the shift register are used to provide L · n registers (for example, 412-0-1, 412-0-L) In order to store the coefficients ps (i), 3 · PS (i) separately. In addition, multiplexers (for example, 413-1, and 415-1) output appropriate coefficient pairs, and this coefficient pair (depending on the input sampling value t-s X (• i)).

L 每一個模組420-0,420-1,.,.,420-R將收到一個 t-s 輸入取樣値X( -i)的一個相對應鄰近位元對L Each module 420-0, 420-1, ..., 420-R will receive a corresponding adjacent bit pair of t-s input sampling value X (-i)

L 經濟部中央標準局貝工消費合作社印製 X2 r,r + 1 (模組420-0收到位元Χ〇 ,Χι ,而 模組420-1收到位元X2 ,X3等等)。在每一個模組中, 例如,模組420-0,其輸入的位元對X〇 ,Xi將在一個 解碼器421中被解碼。這些被解碼後的位元接著將被當 作控制訊號(selector control _signals)般地輸入至η 個多工器 422-0,422-1,...,422-(11-1)。對第1_個多 工器(其中〇SiSn-l) 422_i,接收到一個相對應的係數 對 Ps (i),3· PS (i),而且選擇 0, Ps (i),2· Ps (i) 3*PS (i)其中的一個値(視解碼後的取樣位元而定)以 便產生部份乘積: 2 · r+i Σ P s (i) Χχη ( χη=2 · r t-s -i) I--.--^--1 -裝-----.丨訂-----ί 線 (請先聞讀背面之注意事項再填寫本頁)L The Ministry of Economic Affairs Central Standards Bureau Beigong Consumer Cooperative printed X2 r, r + 1 (module 420-0 received bits X0, X1, and module 420-1 received bits X2, X3, etc.). In each module, for example, module 420-0, its input bit pair X0, Xi will be decoded in a decoder 421. These decoded bits are then input as n control multiplexers 422-0, 422-1, ..., 422- (11-1) like a control signal (selector control_signals). For the 1_th multiplexer (of which 〇SiSn-l) 422_i, a corresponding coefficient pair Ps (i), 3 · PS (i) is received, and 0, Ps (i), 2 · Ps ( i) 3 * PS (i) one of the values (depending on the decoded sampling bits) in order to produce a partial product: 2 · r + i Σ P s (i) Χχη (χη = 2 · r ts -i ) I --.-- ^-1 -installed -----. 丨 order ----- ί line (please read the precautions on the back before filling this page)

L 本紙張尺度適用中國國家標準(CNS ) A4規格(210X297公釐) 83. 3.10,000 五、發明説明(|〇 ) A7 B7 經濟部中央揉準局員工消費合作杜印製 此η個部分乘積將從422_0至422-(n-l)多工器輸出 至用來加總部分乘積的一個相加級(summation stage), 424,以產生如下的和: n-i 2 · r+i t-s fr = Σ [ Σ Ps (i) · Xm ( i=0 m=2 · r 每一個從乘法器模組420-0,420-1,. . .,42 0-R輸出的 和(t),fi (t),...,fR (t)將被複數個加法器相 加,而產出g (t)。而此複數個加法器是以二進位加法樹 (binary adder tree) 430 相連接 〇 每一個前面提及的增頻技巧都有弱點。增頻器200 之所以有弱點,是因爲爲了要求得g (t)必須耗掉許多時 間去執行相乘的動作及加法動作,但加法其耗掉較少程度 的時間。增頻器3〇0之所以有弱點是因爲每—個分接點係 數必須是二次幂的數。此將危害增頻器300的績效。特別 的是,將增頻器3〇〇使用於某種應用領域時,將非常困 難符合某些特定需要的頻率反應。增頻器400的速度很 快,但佔據了 1C晶片太多的表面空間。 這就是爲什麽本發明的目的要提供一個多相位有限脈 衝反應濾波器來克服前述各項技巧的弱點。 本發明之概要説明 本項發明將達成如下的這些及其他目標。根據一項實 施例’父叉對稱排據波器項(cross-symmetric bank filter terms)是被合成,而形成合併濾波器項。假設有 L個濾波器排,被編號爲1,· · .,, . . . , l ,以及在每L This paper scale is applicable to the Chinese National Standard (CNS) A4 specification (210X297mm) 83. 3.10,000 V. Description of invention (| 〇) A7 B7 The Ministry of Economic Affairs Central Bureau of Accreditation and Employee Cooperative Printing of the η partial products The multiplexer from 422_0 to 422- (nl) is output to a summation stage, 424, for summing the partial products to produce the following sum: ni 2 · r + i ts fr = Σ [Σ Ps (i) · Xm (i = 0 m = 2 · r Sum of each output from the multiplier module 420-0, 420-1,..., 42 0-R (t), fi (t), ..., fR (t) will be added by a plurality of adders to produce g (t). The plurality of adders are connected by a binary adder tree 430. The frequency-increasing technique has weaknesses. The reason why the frequency-multiplier 200 has weaknesses is that in order to obtain g (t), it takes a lot of time to perform the multiplication and addition actions, but the addition consumes less. Time. The reason why the frequency multiplier 3000 has a weakness is because each tap point coefficient must be a power of two. This will jeopardize the performance of the frequency multiplier 300. In particular, it will When the frequency multiplier 300 is used in a certain application field, it will be very difficult to meet the frequency response of certain specific needs. The speed of the frequency multiplier 400 is very fast, but it occupies too much surface space on the 1C chip. This is why The purpose of the invention is to provide a multi-phase finite impulse response filter to overcome the weaknesses of the aforementioned techniques. The summary of the present invention illustrates that the present invention will achieve these and other objectives as follows. The cross-symmetric bank filter terms are synthesized to form a merged filter term. Suppose there are L filter rows, numbered 1, · · · · · · · · · l, and in each

----;---:--1 —裝-- (請先聞讀背面之注意事項再填寫本頁) 訂 本紙張尺度逍用中國國家標準(CNS ) A4規格(21〇Χ25&gt;7公釐) 83.3.10,000 ^80794 A7 B7 五、發明説明(11 ) -個濾波器排有幾個被編號爲〇,·.·,i,的分 接點係數。因而,便稱第i排濾波器的第i個濾波器頊 (term)與在第(Li +1)濾波器排中的第(η-i-l)排濾 波器項是交叉對稱。至少兩個交叉對稱濾波器項將被結合 其中濾波器排項如下: 在同相位情況下: t-k----; ---:-1 —installed-- (please read the precautions on the back and then fill in this page) The standard size of the used paper is Chinese National Standard (CNS) A4 specification (21〇Χ25> 7 Mm) 83.3.10,000 ^ 80794 A7 B7 5. Description of the invention (11)-There are several tap coefficients numbered 〇,..., I, in a filter bank. Therefore, it is said that the ith filter term of the ith row filter and the (η-i-1) row filter term in the (Li +1) filter row are cross-symmetric. At least two cross-symmetric filter items will be combined. The filter row items are as follows: In the case of the same phase: t-k

Pk (i) * X( -i),且Pk (i) * X (-i), and

LL

L t-k •(n-l-i))+PL - κ + i (n-l-i) · X ( t-(L-k+l) —(n—1—i)) L」L t-k • (n-l-i)) + PL-κ + i (n-l-i) · X (t- (L-k + l) — (n—1—i)) L ″

L 在非同相位的情況下: 除了乘積項Pl - κ (i) · χ( t_(L-K+l) -i)及L in the case of non-in-phase: except the product term Pl-κ (i) · χ (t_ (L-K + l) -i) and

L ---Γ.--.---If -裝-------訂-----ί 球 (請先聞讀背面之注意事項再填寫本頁) t-k Ρκ (n-l-i)· X( (n-l-i)),的參數i將由 經濟部中央橾準局貝工消費合作社印製 匕 L •1取代外,其餘與同相位情況下相同。 每一個時間t,一個取樣値X( 本紙張尺度逍用中國國家揉準(CNS &gt; Α4規格(210X297公釐) t-kL --- Γ .--.--- If -installed ------- order ----- ί ball (please read the precautions on the back before filling this page) tk Ρκ (nli) · X ((nli)), the parameter i will be replaced by the L · 1 printed by the Ponggong Consumer Cooperative of the Central Bureau of Economics of the Ministry of Economic Affairs, and the rest will be the same as in the same phase. At each time t, a sample value X (the size of the paper is easily calibrated in China (CNS> Α4 specification (210X297mm) t-k

L •i)將被位 83. 3. !〇,〇〇〇 五、發明説明(丨2) A7 B7 移入其中的一個滤波器排,因此改變了其中的排減波器項 而在其他濾波器排的取樣値卻没有改變。定義t=wL+K , 其中1SKSL且其中W是任一非負數的整個0,1,2,..L • i) will be bit 83. 3. 〇〇〇〇〇 5, invention description (丨 2) A7 B7 moved into one of the filter row, so changed the row of the wave reducer item in other filters The sampling value of the row has not changed. Define t = wL + K, where 1SKSL and where W is any non-negative whole 0,1,2, ...

L 當 + 1 SKSL,一個取樣 X( t-k •i)被位移入L when + 1 SKSL, one sample X (t-k • i) is shifted into

L 第K個濾波器排,以至於它是與包含交叉對稱排濾波器 項的第(L-K+1)個濾波器排處在同相位。另外一方面,L The Kth filter row, so that it is in the same phase as the (L-K + 1) th filter row that contains the cross-symmetric row filter term. on the other hand,

L 當-,一個取樣X 2 t-k -i)被位移入第L when-, a sample X 2 t-k -i) is shifted into the first

L K個濾波器排以致於它是與包括交叉對稱排濾波器項的第 (L-K+1)個濾波器排處於非同相位。L K filter rows so that it is not in the same phase as the (L-K + 1) th filter row including cross-symmetric row filter items.

L ----J.--^--1 -裝------訂-----ί 線 (請先聞讀背面之注意事項再填寫本頁) 藉由合併交叉對稱排濾波器項 新的濾波 經濟部中央橾準局貝工消費合作社印製 η 器排將被形成,而此新濾波器排有U= 器項,其中 e 個排濾波 表示e的上限(the ceiling of 本紙張尺度逋用中國國家標準(CNS ) Α4規格(210Χ297公釐) 83. 3.10,000 五、發明説明(丨3) A7 B7 e)或表不小於e+l的最大整數。每—個在時間七及 t-Ι,介於已增頻的輸出値g(t)及g(t-i)之間被改變的 &quot;已合併&quot;(&quot;merged&quot;)濾波器排被定義爲如下的特徵方程 式(characteristic equation)(假設 η 及 L 是偶數): t-k n-i YJ (t)= Σ Pk (i) * X( i=0 其中L ---- J .-- ^-1 -installed ------ order ----- ί line (please read the precautions on the back before filling out this page) By combining cross-symmetric filter New filter items are printed by the Ministry of Economic Affairs, Central Bureau of Industry and Commerce, Beigong Consumer Cooperative. Η device rows will be formed, and this new filter row has U = device items, where e row filters represent the upper limit of e (the ceiling of The paper scale adopts the Chinese National Standard (CNS) Α4 specification (210Χ297mm) 83. 3.10,000 5. Description of the invention (丨 3) A7 B7 e) or the largest integer not less than e + l. Each time-seven and t-Ι, the &quot; merged &quot; (&quot; merged &quot;) filter row that is changed between the increased frequency output values g (t) and g (ti) is defined Is the following characteristic equation (assuming that η and L are even numbers): tk ni YJ (t) = Σ Pk (i) * X (i = 0 where

L -i) + Σ PL - κ + i (i) · X( t-(L-K+l) -i)L -i) + Σ PL-κ + i (i) · X (t- (L-K + l) -i)

L ---Γ--.---f 裝-- (請先閲讀背面之注意事項再填寫本頁) 而且j =k若L --- Γ --.--- f installed-- (please read the notes on the back before filling in this page) and j = kif

L j=L-K+l 若L j = L-K + l if

1 &lt; K &lt; L (9)1 &lt; K &lt; L (9)

L 在 + 1SKSL的情形下,方程式(9)可縮減如下: 經濟部中央標準局貝工消费合作社印製 . υ-ι yj (t)= Σ (Pk (i)+pL - κ i=0 (i)) · [X( t-k t-k -i)+X(L In the case of + 1SKSL, equation (9) can be reduced as follows: Printed by the Beigong Consumer Cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs. Υ-ι yj (t) = Σ (Pk (i) + pL-κ i = 0 ( i)) · (X (tk tk -i) + X (

LL

L 本紙張尺度適用中國國家標準(CNS ) A4規格(210X297公釐) 83.3.10,000 五、發明説明(丨+)L This paper scale is applicable to the Chinese National Standard (CNS) A4 specification (210X297mm) 83.3.10,000 V. Description of the invention (丨 +)

LL

在其中15_KS A7 B7 的情況下,爲了要使大部分的交叉對 稱排濾波器項是處於同相位,非同相位而有著交叉對稱排 濾波器項的濾波器排須能夠被位移。因此,方程式(9)可 被縮減如下: yj (t)= Ρκ (〇) · [X( t-k )+Χ( t+k-1 L L」 L L」 -η))] U-2 Σ (Pk (i)+PL - Κ + 1 (i-l)) * [Χ( t-k t-k i)+X(In the case of 15_KS A7 B7, in order to make most of the cross-symmetric filter entries in the same phase, non-in-phase filter rows with cross-symmetric filter entries must be able to be displaced. Therefore, equation (9) can be reduced as follows: yj (t) = Ρκ (〇) · [X (tk) + Χ (t + k-1 LL "LL" -η))] U-2 Σ (Pk ( i) + PL-Κ + 1 (il)) * (Χ (tk tk i) + X (

LL

t-k • L •(n-i))] 2 · Pk (U-l) · X( (11)t-k • L • (n-i))] 2 · Pk (U-l) · X ((11)

L 簡言之,一個改善後的FIR濾波增頻器具有藉由合 併交叉對稱排濾波器項形成合併濾波器排。此一合併的濾 波器排可大量減少處理時間及所需的電路規模大小(size)。 經濟部中央標準局員工消費合作社印裝 -- ^^1 * ' n .13 ^^1 ^ I I - I...... - ^1 ------- -I- i-4·球 - I I -.......- -- - J. I I I.....- 11- I (請先閲讀背面之注$項再填寫本頁) 本紙張尺度適用中國國家標準(CNS ) A4規格(210X297公釐) 83. 3.10,000 衫0794 五、發明説明(丨5) 經濟部中央標準局貝工消費合作社印製 圖式之簡要説明 第一圖係第一個傳統FiR濾波增頻器 第二(a)—二(f)圖描述了取樣/保持資料位移入第一 圖中FIR濾波增頻器的情形。 第三圖係第二個傳統FIR濾波增頻器。 第四圖係第三個傳統FIR濾波增頻器。 第五圖係第四個傳統FIR濾波增頻器。 第六(a) —六(b)圖係根據本發明,交叉對稱濾波器排 項的合併。 第七圖係根據本發明第一個實施例的一個FIR濾波增 頻器。 第八圖係一個增頻器的功能,並從中選擇分接點係數。 第九圖之(a)-(z),(aa),(bb),(cc)和(dd)描述 了第七圖中FIR濾波增頻器内計數器的狀態。 第十圖係根據本發明第二個實施例的一個FIR濾波 增頻器。 第十一圖係第十圖的FIR濾波器的一個模組核心 (module core) ° 本發明之詳細説明 第六(a)-六(b)係根據本發明交叉對稱排濾波器項的 合併。第六(a)圖係相對於相位12,3,4及5之l=5濾 波器排510,520,530,540及550,且每一個濾波器排具 有π=5分接點係數的一個例子。滤波器排510,5 2 0,5 3 0 , 540及550均有相對應的特徵方程式yi (t) , y2 (t), Y3 (t),y4 (t)及y5 (t),當這些方程式相加在一起時, I----------f -裝------I訂-----^4 (請先閲讀背面之注#^項再填寫本頁) 本紙張尺度適用中國國家棣準(CNS ) A4規格(210X297公釐) 83. 3.10,000 A7 ______B7 五、發明説明(丨6 )In short, an improved FIR filter multiplier has a combined filter bank formed by combining cross-symmetric filter banks. This combined filter bank can greatly reduce the processing time and required circuit size. Printed by the Staff Consumer Cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs-^^ 1 * 'n .13 ^^ 1 ^ II-I ......-^ 1 ------- -I- i-4 · Ball-II -.......---J. II I .....- 11- I (please read the $ item on the back and then fill in this page) CNS) A4 specification (210X297mm) 83. 3.10,000 shirts 0794 V. Description of invention (丨 5) Brief description of the printed patterns of the Beigong Consumer Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs The first figure is the first traditional FiR filter The second (a) -two (f) diagram of the multiplier describes the case where the sample / hold data is shifted into the FIR filter multiplier in the first diagram. The third diagram is the second traditional FIR filter frequency multiplier. The fourth diagram is the third traditional FIR filter frequency multiplier. The fifth diagram is the fourth traditional FIR filter frequency multiplier. The sixth (a) -six (b) diagram is a combination of cross-symmetric filter line items according to the present invention. The seventh figure is a FIR filter multiplier according to the first embodiment of the present invention. Figure 8 shows the function of a frequency multiplier, and selects the tap coefficient from it. (A)-(z), (aa), (bb), (cc) and (dd) of the ninth figure describe the state of the counter in the FIR filter upconverter in the seventh figure. The tenth figure is a FIR filter frequency multiplier according to the second embodiment of the present invention. The eleventh figure is a module core of the FIR filter of the tenth figure. Detailed description of the present invention The sixth (a) -six (b) is the combination of the cross-symmetric row filter items according to the present invention. The sixth (a) graph is an example of l = 5 filter rows 510, 520, 530, 540, and 550 with respect to the phases 12, 3, 4, and 5, and each filter row has a π = 5 tap coefficient. The filter banks 510, 5 2 0, 5 3 0, 540 and 550 have corresponding characteristic equations yi (t), y2 (t), Y3 (t), y4 (t) and y5 (t), when these When the equations are added together, I ---------- f -install ------ I order ----- ^ 4 (please read the note # ^ on the back before filling this page ) This paper scale is applicable to China National Standard (CNS) A4 specification (210X297mm) 83. 3.10,000 A7 ______B7 V. Description of invention (丨 6)

經濟部中央標準局員工消費合作社印製 就可產生增頻器產出値g (t)。在時間t —個取樣輸入値 χ(2)被輸入至濾波器54〇的分接點陣列元(the tap array element) 541,此時 k=4 且 t=wL+K=wL+4 (其中 w 是任何一個非負數的整數〇,l,2,...,且1SK$_L,例如, t=4,9 , I4等等)。(在時問t-Ι,原本在分接點陣列元54 ;l 的取樣値X (1)被位移至分接點陣列元542,而在時間 t-Ι,原本在分接點陣列元542的取樣値χ(〇)被位移至 分接點陣列元543,等等。)。濾波器排510與濾波器排 550處於非同相位。然而,濾波器排520與濾波器排540 則處於同相位。 第六(b)圖係濾波器排項之問的交叉對稱情況。一般 而言,第K個濾波器排的第i個分接點係數的排濾波器 項是與第(L.-K+1)個戚波器排的第(η - i -1)個分接點係數丨 的排濾波器項交叉對稱。排濾波器520和540的排濾波 器項是交叉對稱如下:排濾波器項P2 (〇) · X (2)與項 P4 (4) · X(-2)是交叉對稱的。項P2 (1) · X(l)與項 P4 (3) . X(-l)是交叉對稱。項P2 (2) . X(0)與項 P4 (2) ♦ X(0)是交叉對稱。項P2 (3)。X(-l)與項 P4 (1) · X(l)是交叉對稱的。項P2 (4) · X(-2)與項 Ρ4 (0)·Χ(2)是交叉對稱的。 濾波器排530的排濾波器項是彼此交叉對稱的,項 Ρ3 (〇) · Χ(2)與項Ρ3 (4) · Χ(-2)是交叉對稱的,項 Ρ3 (1) ♦ X(l)與項Ρ3 (3) · X(-l)是交叉對稱的,項 Ρ3 (2) · Χ(0)與本身是交叉對稱的。 濾波器排510和550是處於非同相位。如第六(b) 本紙張尺度適用中國國家榇準_( CNS ) A4规格(210X297公釐) (請先閲讀背面之注意事項再填寫本頁) -----.--------------._-----—iT--- 五、發明説明(丨7 ) A7 B7 圖m示,藉移動減波备排5 與編梃此㈣禎餘二個道t Μ:」 兩個徘濾波鸯磚將蜱捽疋辫R。項5⑴ΛΛ(Η汽I二;二 ?5由· Χ(-2)是交久對稱取。(队♦;奶 p5 ( 2 &gt; · X (-从晕交XM稱的J爲為(;3仏气(_,惠赛.3 p5⑴· x(〇) .是交叉對稱良p齋為…切U(,获專放喊、 ., .. .' . · .. r S , · · · ... - - , - · - · &quot; ' _··.. ·.Ρ5 (0)·Χ(1)是交叉對稱的。項 ρι (0),χίϋ:ν!與項 Ρ5 (4) · Χ(-3&gt;是耷黑舞褥的…:tPrinted by the Employee Consumer Cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs can produce the frequency increaser output value g (t). At time t, a sampling input value χ (2) is input to the tap array element 541 of the filter 54. At this time, k = 4 and t = wL + K = wL + 4 (where w is any non-negative integer 〇, 1, 2, ..., and 1SK $ _L, for example, t = 4,9, I4, etc.). (At time t-Ι, the sampling value X (1) originally at the tap array element 54; l is shifted to the tap array element 542, and at the time t-Ι, originally at the tap array element 542 The sampling value χ (〇) is shifted to tap point array element 543, etc.). The filter bank 510 and the filter bank 550 are not in the same phase. However, the filter bank 520 and the filter bank 540 are in the same phase. The sixth (b) figure is the cross symmetry of the filter line item. Generally speaking, the row filter term of the ith tap coefficient of the Kth filter row is the (η-i -1) th branch of the (L.-K + 1) thirteenth waver row The row filter terms of the contact coefficient are cross-symmetric. The row filter terms of the row filters 520 and 540 are cross-symmetric as follows: the row filter terms P2 (O) X (2) and P4 (4) X (-2) are cross-symmetric. The term P2 (1) · X (l) and the term P4 (3). X (-l) are cross-symmetric. The term P2 (2). X (0) and the term P4 (2) ♦ X (0) are cross-symmetric. Item P2 (3). X (-l) and the term P4 (1) · X (l) are cross-symmetric. The term P2 (4) · X (-2) is cross-symmetric with the term P4 (0) · Χ (2). The row filter terms of the filter bank 530 are cross-symmetrical to each other, and the term P3 (〇) · Χ (2) and the term P3 (4) · Χ (-2) are cross-symmetrical, the term P3 (1) ♦ X ( l) is cross-symmetric with the term P3 (3) · X (-l), and the term P3 (2) · X (0) is cross-symmetric with itself. The filter banks 510 and 550 are not in phase. Such as the sixth (b) This paper standard is applicable to the Chinese National Standard _ (CNS) A4 specification (210X297mm) (please read the precautions on the back before filling this page) -----.------ --------._-----— iT --- Fifth, the description of the invention (丨 7) A7 B7 Figure m shows that by mobile wave reduction equipment row 5 and the editing of the remaining two Road t Μ: "Two squid filter duck bricks will tickle R. Item 5 (1) ΛΛ (H steam I two; two? 5 by · Χ (-2) is symmetrical to take long time. (Team ♦; milk p5 (2 &gt; · X (-from the halo XM called J is (; 3 Qi (_, Hui Sai. 3 p5 ⑴ · x (〇). Is a cross-symmetrical good p zai for ... cut U (, get a special shout,., .. ''... R S, · · ·. ..--,-·-&quot; '_ ·· .. ·· P5 (0) · Χ (1) is cross-symmetrical. Item ρι (0), χίϋ: ν! And item Ρ5 (4) · Χ (-3 &gt; is black dance mat ...: t

LL

上述的合併將邊生W ‘値它合併t的%Λ'# h 排,而每一個已合併了的濾波器排將有著υ= 接點陣列元。每對交叉對稱排濾波器項是: 對處於同相位情況: t-k t-(L-K+l) Ρκ (i) · X ( I -I -i) &gt; Pl - κ + i ⑴·ϊί(Ι I -i)The merger described above will generate a row of '% Λ' # h which is merged with t, and each merged filter row will have a ν = contact array element. Each pair of crossed symmetrical filter terms is: For the case of the same phase: tk t- (L-K + l) Ρκ (i) · X (I -I -i) &gt; Pl-κ + i ⑴ · ϊί ( Ι I -i)

LL

L 個分L points

t-k t-(L~K+l) Pk (n-l-i) · X( I -」-(n-l-i&gt;)且 Pl - κ + i (n-l-i) ·)((匕-j J -(n-l-i)) l L 」 L I---^--.---f I裝------訂-----f 球 (請先閲讀背面之注意事項再填寫本頁) 經濟部中央標準局貝工消費合作社印裝 對處於非同相位的情況: 除了乘積項中的i參數,Pl - κ t-(L-K+l) (i)tk t- (L ~ K + l) Pk (nli) · X (I-"-(nl-i>) and Pl-κ + i (nli) ·) ((dagger-j J-(nli)) l L ”L I --- ^ --.--- f I installed ------ order ----- f ball (please read the precautions on the back before filling in this page) Central Ministry of Economic Affairs The printing and printing pair of the industrial and consumer cooperative is in the same phase: except for the i parameter in the product term, Pl-κ t- (L-K + l) (i)

t-K X(t-K X (

L -i)及 PK (n-l-i) · X( L 」 本紙張尺度適用中國國家標準(CNS ) A4規格(210X297公釐) 83. 3. 10,000 五、發明説明(丨牙) A7 B7 (n-1-i))將被i-i取代外,其他如同相位的情況中》 使用上述配對的交叉對稱排濾波器項合併方式,每 個合併滤波器排.的增頻特徵’(interpolation . .. . · ... . characteristic) yj (t)產生如下(假設L是偶數): y Π-1 yj (t)=E pr (i) · χ( t-k n-1 -i)+E Pl - κ + 1 (i) * X( t-(L-K+l) -i) (9)L -i) and PK (nli) · X (L) This paper scale is applicable to the Chinese National Standard (CNS) A4 specification (210X297 mm) 83. 3. 10,000 V. Description of invention (丨 tooth) A7 B7 (n-1 -i)) will be replaced by ii, in other cases like phases, the above-mentioned paired cross-symmetric row filter item combination method is used, and the frequency-increasing feature of each combined filter row is' (interpolation ... ... characteristic) yj (t) produces the following (assuming L is even): y Π-1 yj (t) = E pr (i) · χ (tk n-1 -i) + E Pl-κ + 1 ( i) * X (t- (L-K + l) -i) (9)

LL

L t-k 對每一個時間t,一個取樣値x( -i)被L t-k For each time t, a sample value x (-i) is

L 經濟部中央標準局員工消費合作社印製 位移入一個增頻濾波器500的相對應的第K個濾器排 51〇-55〇。一般而言,這會改變其中一個濾波器排510, 520,530,540或550的値。相同地,對每一個t,只 有已合併了的交叉對稱排濾波器項的一個已合併了的濾波 器排將改變其輸出値。因此,爲求得任一時間t下的 g(t)値,已被改變的合併濾波器排輸出値y]必須使用 方程式(9)來決定。這個已被改變了的輸出値因此被加至 先前已被決定之已合併的濾波器排輸出値y] (t),而其 値在t-i與t之間尚未改變(假如有的話),(如下所討 I&quot;丨J--:---^ I裝-----^-丨訂-----一 或 (請先聞讀背面之注意事項再填寫本頁) 本紙張尺度適用中國國家揉準(CNS ) A4規格(210X297公釐) 83. 3.10,000 A7 B7 五、發明説明(丨f 論,對L &gt; 2的相位,在每一時間t,總存在著—個不 會改變其値的已合併的濾波器排)。 方程式(9)可被更進一步地筒化,但是,簡化的程 度端視收到下一個取樣値所處的時間。詳細地説,若Printed by the Employee Consumer Cooperative of the Central Bureau of Standards of the Ministry of Economy Shifted into a corresponding K-th filter row 51〇-55〇 of an up-frequency filter 500. Generally speaking, this will change the value of one of the filter banks 510, 520, 530, 540 or 550. Similarly, for each t, only a merged filter bank that has merged cross-symmetric bank filter terms will change its output value. Therefore, in order to find the value of g (t) at any time t, the output value of the combined filter bank that has been changed] must be determined using equation (9). This changed output value is therefore added to the output of the combined filter row that has been previously determined]] (t), and its value has not changed between ti and t (if any), ( As discussed below I &quot; 丨 J-: --- ^ I installed ----- ^-丨 order ----- one or (please read the precautions on the back and then fill out this page) This paper size is applicable China National Standard (CNS) A4 specification (210X297mm) 83. 3.10,000 A7 B7 V. Description of invention (丨 f theory, for the phase of L &gt; 2, at each time t, there is always a- (It will change the value of the combined filter bank). Equation (9) can be further simplified, but the degree of simplification depends on the time when the next sample value is received. In detail, if

L (1+ -) S K S L (其中,則改變其値的合 2 併濾波器排僅僅只是組合同相位的交叉對稱排濾波器項。 在(1 +L (1+-) S K S L (where, change the value of the combination 2 and the filter bank is just to combine the cross-symmetrical filter terms of the same phase. In (1 +

L (請先閲讀背面之注意事項再填寫本萸) 經濟部中央標隼局員工消費合作社印製 t-k wL+K-K _ Li - — L - t+k-1. wL+K+K-1 — L· - ,若(1 + L ) SK£L,則 2 t-k t+k-i —L - _ L -L (Please read the precautions on the back before filling in this book) Printed by the Ministry of Economic Affairs, Central Standard Falcon Bureau Employee Consumer Cooperatives tk wL + KK _ Li-— L-t + k-1. WL + K + K-1 — L ·-, If (1 + L) SK £ L, then 2 tk t + ki —L-_ L-

W W+ 訂· 2K-1 (12) W+1 方程式(9)可被重寫如下(假設L及η均爲偶數) ' u-l t-k n-l t-k yj (t) = Z PK (i) . χ( I ——I -i)*£ PK (i) · X( i=0 L L J i=|j L~J'n U-l Σ i=0 fK-i -K ♦ t (i) ·Χ(W W + 2K-1 (12) W + 1 Equation (9) can be rewritten as follows (assuming L and η are even numbers) 'ul tk nl tk yj (t) = Z PK (i). Χ (I ——I -i) * £ PK (i) · X (i = 0 LLJ i = | j L ~ J'n Ul Σ i = 0 fK-i -K ♦ t (i) · Χ (

LTLT

n-l -i-n*z i=U t*K-l X( !-I -i-i) (13) 本紙張尺度逋用中國國家標準(CNS ) A4规格(21〇Χ:297公釐) 五、發明説明(Μ A7 B7 重組方程(I2)及(I3)則產生 〆 U-1 t-k y〇 (t)=Z (Pk (i) # +Pl - k + x (i)) · [X( i=onl -in * zi = U t * Kl X (! -I -ii) (13) This paper uses the Chinese National Standard (CNS) A4 specification (21〇Χ: 297 mm) V. Description of the invention (Μ A7 B7 Reorganization equations (I2) and (I3) produce 〆U-1 tk y〇 (t) = Z (Pk (i) # + Pl-k + x (i)) · [X (i = o

L i)· t-k X( 在 1&lt;K&lt;-L i) t-k X (at 1 &lt; K &lt;-

L 2 t-k L L t+k-iL 2 t-k L L t + k-i

L 的情況下:In the case of L:

wL+K—KwL + K—K

=W (請先閱讀背面之注意事項再填寫本頁) ,衣.= W (please read the notes on the back before filling out this page), clothing.

L 2K-1 經濟部中央標準局員工消費合作杜印製 wL+K+K—1 — L — — L 一 因此 -若 1&lt;Κ&lt; L 則 2 t-k t+K-l 一 L - — L - 組合方式式(13)及 (15) 則產生 yj = Ρκ (〇) · [χ( t-k )+X( t+k -1 -L - -L - -n)] u_2 t-k Σ (PK (i)+Pl - κ + i (i-l)) · [X( i=l 2PK (U-l) · X(L 2K-1 The Ministry of Economic Affairs Central Standards Bureau employee consumption cooperation du printing wL + K + K—1 — L — — L-therefore-if 1 &lt; Κ &lt; L, then 2 tk t + Kl-L-— L-combination Equations (13) and (15) yield yj = Ρκ (〇) · [χ (tk) + X (t + k -1 -L--L--n)] u_2 tk Σ (PK (i) + Pl -κ + i (il)) · (X (i = l 2PK (Ul) · X (

=W= W

L (15) -i)+X( 訂 L」 t-kL (15) -i) + X (order L '' t-k

L t-k -W-l)L t-k -W-l)

L (16) 本紙張尺度適用中國國家標準(CNS ) A4規格(210X297公釐) 3总〇794 A7 B7 經濟部中央標準局貝工消費合作社印製 五、發明説明(21 ) 在方程式(I6)中,第一及最後加數的產生是由於排濾波器 項處於非同相位(這些項不具有交叉對稱排濾波器項)。 當任一η和L均非偶數時,方程式(14)及(16) 將被修改成具有不同加數。第七圖描述了根據本發明的一 個實施例的一個FIR濾波器增頻電路600。增頻器600 有一個隨機存取記憶體(RAM) 610用來接收及儲存被輸 入的取樣値X t-k X( —* L — t-k -i)至 Χ( t-k 一 L - —L 一 -i-n)的n+1個取樣値的 _i)。RAM 610有用來儲存從 容量。圖解説明了每一個取樣有b個位元。RAM 610在時 問t的每L個相位接收到一個斩的値X( t-k i) 〇L (16) This paper scale applies the Chinese National Standard (CNS) A4 specification (210X297 mm) 3 Total 〇794 A7 B7 Printed by the Beigong Consumer Cooperative of the Central Bureau of Standards of the Ministry of Economy V. Description of invention (21) In equation (I6) In, the first and last addends are generated because the row filter terms are not in phase (these terms do not have cross-symmetric row filter terms). When either η and L are not even numbers, equations (14) and (16) will be modified to have different addends. The seventh figure depicts a FIR filter frequency increasing circuit 600 according to an embodiment of the present invention. The frequency multiplier 600 has a random access memory (RAM) 610 for receiving and storing the input sample values X tk X (— * L — tk -i) to Χ (tk -L--L --in) N + 1 sampling values of _i). RAM 610 is used to store secondary capacity. The illustration shows that each sample has b bits. The RAM 610 receives a chopped value X (t-k i) for every L phases of time t.

L 一個位址產生器620被連接至隨機存取記憶體610。 此位址產生器包括一個可裝塡的(loadable)上計數器 (up-counter) 622及一個可裝塡的下計數器(down-counter) 624 。 一個上計數鎖存器 (up-counter latch) 621被連接至多工器626的輸出値且鎖存器621的輸出 値被接至上計數器622的一個輸入値。一個下計數鎖存 I Μ------&lt; ·裝-----^--訂----- (請先閲讀背面之注意事項再填寫本頁) 本纸張尺度適用中國國家標準(CNS ) Α4規格(210X297公釐) 83. 3.10,000 五、發明説明(22) A7 B7 經濟部中央標準局貝工消费合作社印裝 器 (down-counter latch) 623被連接至下計數器624 的輸出値,且鎖存器623的輸出値被接至下計數器624 的一個輸入値,一個多工器626被提供用來選擇上計數器 622的輸出値或下計數器624的輸出値,端視一個外在控 制訊號。多工器626的輸出値被輸出至隨機存取記憶體 610的位址埠(address port) °圖解説明了 一個二相位鐘 (two phase clock)被用來使上計數器622可以往上計數 及使下計數器624可以往下計數。 隨機存取記憶體610的輸出値被連接至一個鎖存器 612及一個加法器614的輸入値。鎖存器612的輸出値 被連接至加法器614的第二個輸入値。在隨機存取記憶 體610中被位址產生器620定了位址的取樣,可以被選 擇性地被輸出至加法器614或鎖存器612。 —個唯讀記慷體630被用來儲存每一個已合併的排 濾波器項的分接點係數(例如,交叉對稱排濾波器項的分 接點係數和(sum)及非同相位合併排濾波器項的兩個分 接點係數)。唯讀記憶體630必須要有一個c.d的容量 (capacity),其中分接點係數可能會有一個d位元的精準 n L 度(d bit precision) 及其中 C=L · - + -〇 對 n=4 2 2 且L=4而言,唯讀記憶體630有一個10d的最小容量。 然而,一個容量14d的唯讀記憶體被用來做説明。額外的 容量是被用來儲存分隔相對應於每一個已合併的濾波器排 的分接點係數値的零値分隔器(0 value separators)。這 (請先閲讀背面之注意事項再填寫本頁) 本紙張尺度適用中國國家標準(CNS ) A4規格(210x297公釐) 83. 3.10,000 五、發明説明(23 ) A7 B7 經濟部中央橾準局員工消费合作社印製 樣可以簡化控制。説明如下。唯讀記憶體63〇的位址產生 器可以是一個簡單的上計數器632。 被唯讀記憶體計數器(ROM counter) 632定位址的分 接點係數及其和將由唯讀記憶體63 〇輸出至一個乘法器 64〇。由加法器614輸出的取樣和亦將當作乘法器640的 輸入。乘法器64〇將由加法器614輸出的每一個取樣値 的和與由唯讀記憶體63〇輸出之其相對應的分接點係數値 相乘。因此形成的乘積是—個已合併的排濾波器項上,且 自乘法器64〇輸出至一個多工器642。多工器642的輸出値 連接至累計器644。累計器644有一個輸出値其被視爲第 二個輸入値般地被重新接至累計器644。使用多工器642 與累計器644之間互相連接才有可能將一系列從乘法器 64〇輸出的乘積相加。例如,合併的排濾波器項,以產出 合併濾波器排的輸出y] (t)。最後,累計器644將被首 先歸零(reset)。然後,一個濾波器排加總控制訊號(由 一個没被列出的控制電路所產生)使多工器642選擇由 乘法器64〇輸出的乘積。這些乘積藉由自累計器644的 輸出至累器644的第二個輸入値的回饋連接(feedback connection)而被一次 (at a time)加總至現存和中,該 現存和存於累計器644中。此外,累計器的輸出値被連接 至一個鎖存器646。此鎖存器的輸出値被視爲第二個輸入 値般被連接至多工器642。在累計器644中形成的一個 已合併的濾波器排輸出値y〗(t)可被儲存在鎖存器646 I----^------f 袭-- (請先聞讀背面之注意事項再填寫本頁) 訂 本紙張尺度適用中國國家標準(CNS ) A4規格(210X297公釐) 83. 3.10,000 A7 __^_____ B7 五、發明説明(2斗) 中。在下一個時間t+i,此已合併的濾波器排輸出値 yj (t)可被加至在累計器644中形成的另外一 個已合併的濾波器排輸出値y〗(t+1)。 對每一個濾波器排有n=4個分接點係數及其L=4個 濾波器排,一共會有16個分接點係數Pl (〇),...,L An address generator 620 is connected to the random access memory 610. The address generator includes a loadable up-counter 622 and a loadable down-counter 624. An up-counter latch 621 is connected to the output value of the multiplexer 626 and the output value of the latch 621 is connected to an input value of the up-counter 622. A lower count latch I Μ ------ &lt; · installation ----- ^-order ----- (please read the precautions on the back before filling in this page) This paper size is applicable to China National Standard (CNS) Α4 specification (210X297mm) 83. 3.10,000 V. Description of the invention (22) A7 B7 The Down-counter Latch 623 of the Beigong Consumer Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs is connected to the down counter The output value of 624, and the output value of latch 623 is connected to an input value of down counter 624, a multiplexer 626 is provided to select the output value of up counter 622 or the output value of down counter 624, depending on An external control signal. The output value of the multiplexer 626 is output to the address port of the random access memory 610. The diagram illustrates that a two phase clock is used to enable the up counter 622 to count up and use The down counter 624 can count down. The output value of the random access memory 610 is connected to the input value of a latch 612 and an adder 614. The output value of the latch 612 is connected to the second input value of the adder 614. The samples addressed by the address generator 620 in the random access memory 610 can be selectively output to the adder 614 or the latch 612. A read-only memory 630 is used to store the tap coefficients of each combined row filter term (eg, the sum of the tap coefficients of the cross-symmetric row filter term and the non-in-phase combined rank The two tap coefficients of the filter term). The read-only memory 630 must have a cd capacity, in which the tap coefficient may have a d bit precision n L degree (d bit precision) and C = L ·-+ -〇 to n = 4 2 2 and L = 4, the read-only memory 630 has a minimum capacity of 10 days. However, a read-only memory with a capacity of 14d was used for illustration. The extra capacity is used to store zero value separators that separate the tap coefficients corresponding to each combined filter bank. This (please read the precautions on the back before filling out this page) The paper size is applicable to the Chinese National Standard (CNS) A4 specification (210x297mm) 83. 3.10,000 V. Description of invention (23) A7 B7 Central Ministry of Economic Affairs The printing of samples by the Bureau ’s consumer cooperatives can simplify control. described as follows. The address generator of the read-only memory 63 may be a simple up counter 632. The tap coefficient and the sum of the addresses addressed by the ROM counter 632 will be output from the ROM 63 〇 to a multiplier 64 〇. The sample sum output by the adder 614 will also be used as the input of the multiplier 640. The multiplier 64o multiplies the sum of each sample value output by the adder 614 and its corresponding tap point coefficient value output by the read-only memory 63o. The product thus formed is a combined row filter term, and is output from a multiplier 640 to a multiplexer 642. The output value of the multiplexer 642 is connected to the accumulator 644. The accumulator 644 has an output value which is regarded as the second input and is reconnected to the accumulator 644 as usual. It is possible to add a series of products output from the multiplier 64 using the multiplexer 642 and the accumulator 644 connected to each other. For example, the combined bank filter term can produce the output y] (t) of the combined bank. Finally, the accumulator 644 will be reset first. Then, a filter bank sums up the total control signal (generated by a control circuit not listed) to cause the multiplexer 642 to select the product output by the multiplier 64. These products are added to the existing sum at a time by the feedback connection from the output of the accumulator 644 to the second input value of the accumulator 644. The existing sum is stored in the accumulator 644 in. In addition, the output value of the accumulator is connected to a latch 646. The output value of this latch is regarded as the second input and is generally connected to the multiplexer 642. The output value of a combined filter bank formed in the accumulator 644 (t) can be stored in the latch 646 I ---- ^ ------ f strike-- (please read first Note on the back and then fill out this page) The size of the revised paper is in accordance with the Chinese National Standard (CNS) A4 specification (210X297mm) 83. 3.10,000 A7 __ ^ _____ B7 5. Description of the invention (2 buckets). At the next time t + i, this combined filter bank output value yj (t) can be added to another combined filter bank output value formed in the accumulator 644 (t + 1). For each filter bank there are n = 4 tap coefficients and its L = 4 filter banks, there will be a total of 16 tap coefficients Pl (〇), ...,

Pi (.3),P2 (〇),. · .,P2 (3),. . .,P4 (3)。然而,一 共只有8個獨特的分接點係數値hD -h7 ,如同第八圖 所示的一個功能中取出。這些獨特的係數値hD -h7被指 派至任一個濾波器排並摘要如下表一所示: n n — J n n I I I (請先閲讀背面之注意事項再填寫本頁) -訂- 表一 値 h〇 hi h2 h3 h4 h5 he h? 分接點 Pi (〇) P2 (0) P3 (0) P4 (0) Pi (i) p2 (i) p3 (1) p4 (1) 分接點 P4 (3) P3 (3) P2 (3) Pi (3) P4 (2) P3 (2) p2 (2) Pi (2) 經濟部中央標準局貝工消費合作社印製 在將這些分接點係數hQ -h7代入方程式(11)及(I2) 後,在每一個t及t-i之間已合併的濾波器排特徵將會 改變,如下描述: 本紙張尺度適用中國國家榡準(CNS ) A4規格(210X297公釐) 83.3. !〇,〇〇〇 3^〇794 A7 ________B7 五、發明説明(25*) 表 在t輿t-i之間 改铤了的合併了的濾波器排 t=l 典 t=0 y! (1)= yi (l&gt;+y4 (l)=h〇 (x(0)+x(-4))+(h3 +h4 )(χ(-1)+χ(-3)) ♦2h7 x(-2) t=2 典 t=l y2 (2)= y2 (2)*y3 (2)=h1 (x (0)*x(-4)) + (h2 +h5 )(x(-1)+x(-3)) ♦2ht x(-2) t=3 輿 t=2 yz (3)= y2 (3)*ya (3)=(hi +hz )(x(0)+x(-3))+(hs *h4 )(x(-1)*x(-2)) t=4 典 t=3 yi (4)= yi (4)*y4 (4)=(h〇 +h3 )(x(0)*x(-3))+(h4 *h7 )(x(-1)*x(-2)) t=5 典 yi (5)= yL (5)+y4 (5)=ha (x(1)+x(-3))+(h3 +h4 )(x(0)*x(-2))+ 2h? x(-l) t=6 與 t=5 y2 (6)= y2 (6)+ya (6)=hl (x(1)+x(-3)) + (h2 +h5 )(x(0)+x(-2))♦ 2h&amp; x(-l) t=7 與 t=6 yi (7)= yz (7)+y3 (7) = (ht +h2 ) (x(1)+x(-2))+ (h5 +h6 ) (x(0)+x(&quot;1)&gt; ----;-----ί 威------ (請先聞讀背面之注意事項再填寫本貰) 經濟部中央標準局貝工消費合作社印製 可能較有價値的是,g (t)可將在t-i及t之間會改變 値的已合併濾波器排輸出値y] (t)加至在t-l及t之間 不會改變値的已合併濾波器排輸出値yi (t),y〖(t), 83. 3.10,000 本紙張尺度適用中國國家標準(CNS ) Α4規格(210X297公釐) A7 ________B7五、發明説明(26) 經濟部中央標準局員工消费合作社印裝 …,yj - 1 (t) » yj + 1 (t),yj + 2 (t),..., y^J (t)而被求得。在L=4的情況下,僅僅只有二個合併 濾波器排。那就是,一個在t-Ι及t之間會改變値的濾 波器排輸出値(t)及一個在t-Ι及t之間不會改變 値的濾波器排輸出値y] (t)。 現在來討論關於第九圖(a)-(z),(aa),(bb),(cc) 及(dd)在L==4及N=4情況下本發明運作的情形◊首先, 第九圖描述的是計數器622 , 624及632 ,隨機存取記憶 體610及唯讀記憶體630的狀態。唯讀記憶體有14個儲存 格(storage cell)用來儲存分接點係數及交叉對稱排滤 波器項的分接點係數的和(sums)。根據表二所示的已改 變的已合併濾波器排,分接點係數或分接點係數和將被組 合(grouped)。此組合被零値(0 values)所區隔 (separated )。隨機存取記憶體610有n+1或5個儲 存格首先用來儲存5個取樣値x(〇),x(-l),x(-2), x(-3)及x(-4)。對每時間t的L週期(L time t-k periods),一個新的取樣値X( I - -i)被位移入用 來儲存最早之取樣値的隨機存取記憶體610的儲存格中。 ----:--:---ί 裝-----^-丨訂-----ί 鉍 (請先閲讀背面之注意事項再填寫本頁) 本紙張尺度適用中國國家揉準(CNS ) Α4規格(210X297公釐) 83. 3.10,000 3 20794 A7 B7 五、發明説明(27 ) 經濟部中央標準局貝工消费合作社印装 唯讀記憶體計數器632被設初値點至唯讀記憶體630的 第1格(the first cell)(例如,設定至〇),上計數器 622被設初値點至隨機存取記憶體610的第1格(例如, 設定至0)且下計數器624被設初値點至隨機存取記憶 體610的最後1格(the last cell)(例如,設定至4)〇 上計數鎖存器621及下計數鎖存器623分別被設初値爲 0與4 〇 在第一個脈衝鐘(a first clock pulse)的第一個相 位之時第九(a)圖,多工器626選擇了上計數器622的計 數來定隨機存取記憶體610的位址。隨機存取記憶體61〇 因此輸出由上計數器622所指出的取樣値x(0)。取樣値 x(〇)被儲存在鎖存器612之中。 在第一個脈衝鐘的第二個相位之時(第九(b)圖),上 計數器622上計數至1。多工器626選擇下計數器624 的數來定隨機存取記憶體610的位址。回應地,隨機存 取記憶體610輸出取樣値x(_4)。加法器614將取樣値 x(_4)與存在鎖存器612中的取樣値x(〇)相加◊加法 器614輸出和x(〇)+x(_4)至乘法器64〇。進一步地, 唯讀記憶體63〇將由唯讀記憶體計數器632定過位址的 h〇値輸出至乘法器640。乘法640將h〇與和x(〇) + Χ(_4)相乘且將乘積(例如,排濾波器項)hD (Χ(0) + x(-4))輸出至多工器642。此多工器將乘積h0 (Χ(0) + x(-4))輸出至累計器644。 在第二個循環的第一個相位9 (第九(c)圖),多工器 626選擇上計數器622的數來爲隨機存取記憶體610定 ----.-----f 裝— (請先閲讀背面之注意事項再填寫本頁) 本紙張尺度逋用中國國家標準(CNS ) A4規格(210X297公釐) 83. 3.10,000 經濟部中央梯準局負工消費合作社印裝 妓 07 94 A7 __ B7 五、發明説明(2分) 位址。回應地,隨機存取記憶體610將取樣値x(-l)輸 出儲存於鎖存器612中。同時地,下計數器下計數至3 且唯讀記憶體計數器632上計數至1。 在第二個循環的声二個相位上(第九(d)圖),多工器 626選擇下計數器624的數來爲隨機存取記憶體610來 定位址。回應地,隨機存取記憶體610將取樣値χ(-3) 輸出至加法器614,在那兒與鎖存器612中的取樣値 X(-l)相加。同時地,唯讀記憶體630輸出被計數器632 定位址的分接點係數h3 +h4的組合。和h3 +h4與和 .X(-l)+X(_3)在乘法器64〇中相乘且因而所形成的乘積 (h3 +h4 )(X(-l)+x(-3))被輸出至多工器642。多工器 642選擇自乘法器640輸出的乘積(h3 +h4 )(χ(-1) + Χ(-3))做爲輸入累計器644的輸入値。累計器644將 乘積(h3 +h4 )(χ(-1)+χ(_3))加至已經儲存在那褢的乘 積 h〇 (Χ(0)+Χ(-4)) ° 在第三個循環的第一個相位上,(第九(e)圖),多工 器626選擇上計數器622的數來爲隨機存取記憶體61〇 定位址。回應地,隨機存取記憶體610將取樣値χ(-2) 輸出儲存於鎖存器612中。同時地,下計數器下計數至 2且唯讀記憶體計數器632上計數至2。 在第三個循環的第二個相位上第九(f)圖多工器626 選擇下計器624的數來爲隨機存取記憶體610定位址。 隨機存取記憶體610將取樣値x(-2)輸出至加法器614, 在那兒與鎖存器612中的取樣値X (-2)相加,同時地, 唯讀記憶63〇將被唯讀記憶體632定位址的分接點係數 I----------f ·裝------訂----- (請先閲讀背面之注意事項再填寫本頁) 本紙張尺度適用中國國家標準(CNS ) A4規格(210X297公釐) 83. 3.10,000 32〇tg4 A7 ___________B7五、發明説明(2?) 經濟部中夬標準局属工消费合作社印製 h?輸出。分接點係數h7在乘法器640中與2 x(-2) 相乘,且因而形成的乘積2h7 X卜2)被輸出至多工器642。 多工器642選擇由乘法器64〇輸出的乘積2h? Χ(_2)做 爲累計器644的輸入値。累計器644將乘積2h7 χ(-2) 與已存於其中的乘積和(h3 +h4 )(χ(-ι)+χ(-3)) + h0 (x(0)+x(-4))相加。因此,累計器644產生產出物 〆 yi (t) 〇 第四循環的第一相位上(第九(g)圖)上下計數器622 及624將與分別接連其上的鎖存器621及623儲存的數 載入。進一步地,唯讀記憶體計數器632下計數至3。 乘積的和 2h7 x(-2) + (h3 +h4 )(x(-i)+x(-3))+h〇 (x(0) +X(_4))儲存於累計器644中且被輸出至鎖存器636並 儲存在那裏。乘積的代表在t及t+i之間g(t+i)不會 改變的部分(portion)。那是,在t=l及t+l=2之間不 會改變的合併濾波器排輸出値yl (t)。 在第四個循環第二個相位上(第九(h)圖),唯讀記憶 體計數器632爲唯讀記憶體630定位址。然而,被定位 址的値爲零,因而,只有零値從乘法器640輸出。 第五個循環的第一個相位上(第九(i)圖),多工器 626選擇上計數器622的數來爲隨機存取記憶體610定 位址。這使得取樣値X(〇)由隨機存取記憶體610輸出 並儲存於鎖存器612中。唯讀記憶體計數器632上計數 至 4 〇 (請先閲讀背面之注意事項再填寫本頁) 本紙張尺度適用中國國家揉準(CNS ) A4規格(210X297公釐) 83. 3.10,000 A7 ____B7五、發明説明(3〇 ) 經濟部中央標準局貝工消費合作社印製 第五個循環的第二個相位上,(第九(j)圖),多工器 626選擇下計器624的來數爲隨機存取記憶體610定位 址。這使得由隨機存取記憶體610輸出的取樣値輸入至 加法器614。此加法器,將取樣値X(-4)與儲存於鎖存器 612中的取樣値χ(〇〉相加,且將和χ(_4)+χ(0)輸出至 乘法器640。唯讀記憶體計數器632將在那褢的唯讀記憶 體630定位址,使得hi値被輸出至乘法器640。乘法器 640將hi値與和x(-4)+x(0)相乘,且將乘積hi (x(-4) +X(0))輸出至多工器642。多工器642將乘積hi (X(-4) +X(0))輸出至累計器644,並在那裏將乘積加以儲存。 第六個循環的第一個相位上(第九(K)圖),多工器626 選擇上計數器622的數爲隨機存取記憶體610定位址。 這使得取樣値X(-l)由隨機存取記憶體輸出並且儲 存於鎖存器612中。唯讀記憶體計數器632上計數至5。 進一步地,下計數器624下計數至3。 在第六個循環第二個相位上(第九(1)圖),多工器 626選擇下計數器624的數爲隨機存取記憶體610定位 址。這使得取樣値x(_3)由隨機存取記憶體61〇輸出至 加法器614。此加法器將取樣値X(-3)與儲存於鎖存器 612中的取樣値X(-l)相加且將和x(_3)+x(_l)輸出至 乘法器640。唯讀記憶體計數器63 2將在那裏的的唯讀 記憶體630定位址,使得h2 +h5値被輸出至乘法器 640。乘法器640將h2 +h5値與和x(-3)+x(-l)相乘 且將乘積(h2 +h5 )(Χ(-3)+Χ(-1))輸出至多工器642, 多工器642將乘積(h2 +h5 )(x(-3)+x(-l))輸出至累 I------; I :--^ -裝------訂 (請先閲讀背面之注$項再填寫本頁) 本紙張尺度適用中國國家標準(CNS ) A4規格(210X297公釐) 83. 3.10,000 五、發明説明(31 ) A7 B7 經濟部中央揉準局貝工消费合作社印製 計器644,在那裏乘積被相加至已儲存在那褢的乘積 hi (X(-4)+}£(〇))。下計數器634的數被儲存於鎖存器 623 中 ° 第七個循璁的第一個相位上(第九(m)圖)多工器626 選擇上計數器622的數來爲隨機存取記憶體610定位址。 這使得隨機存取記憶體610將已儲存於鎖存器612中的 取樣値X(-2)輸出。唯讀記憶體計數器632上計數至6。 更進一步地,下計數器下計數至2。 第七個循環的第二個相位上(第九(η)圖),多工器 626選擇下計數器624的數來爲隨機存取記憶體610定 位址。這使得隨機存取記憶體610將取樣値χ(-2)輸出 至加法器614。加法器614將取樣値χ(-2)與儲存於鎖 存器612中的取樣値x(-2)相加且將·2Χ(_2)値輸出至 乘法器640。同時,唯讀記憶體計數器632將在那裏的 唯讀記憶體630定位址,使得唯讀記憶體630將h6値 輸出至乘法器64〇,乘法器64〇將h6値與2 x(_2)相 乘且將乘積2h6 X(-2)輸出至多工器642。多工器642 將乘積2h6 X(_2)輸出至累計器644,在那褢。乘積被 加於已儲存在那裏的乘積和(h2 +h5 )(x(-3)+x(-l)) + hi (X(-4)+X(0))。在累計器644所產生的和(h2 +h5 ) (xi-SJ+xi-lH+hi (x(-4)+x(0))+2h6 x(-2)斡是合併濾 波器排yl (t)的輸出値。 第八個循環的第一個相位上(第九(〇)圖),上計數器 622及下計數器624將連接於其上的鎖存器621或623 m. 1^1 m I (請先閲讀背面之注意事項再填寫本頁) -訂 &quot; 本紙張尺度適用中國國家揉準(CNS ) A4規格(210X297公釐) 83. 3.10,000 320794 A7 B7 五、發明説明(多2) 經濟部中央橾準局員工消費合作社印製 分別儲存的數載入。然而與第一個循環的第一個相位(第 九(a)圖)及第四個循環的第一個相位(第九(g)圖)不同, 输入至下計數器624的數是3。(在第6個循環的第2 個相位中,此數已被儲存於鎖存器623中)。唯讀記憶體 計數器632上計數至7。 多工器642選擇相對應於g(2&gt;没有改變的部份的 和作爲累計器644的輸入値。結果,此累計器將未改變 的濾波器排輸出値yi (1)相加至剛被算出來的已改變了 的濾波器排輸出値h (2)。這個相對應於g(2)的和被 輸出至累計器644。 如表二所示,合併濾波器排輸出値(2)在t=2及 t+l=3之間再度改變,例如y2 (3)不等於(2)。然 而,yi (t)在t=2及t + l = 3之間不會改變。因而,合 併濾波器排輸出値 y2 (2)不會儲存於鎖存器 646之中。 反而,鎖存器646保存著(retains)合併濾波器排输出 値Yi (t)以決定g(3)。 :—.---f.-裝-----^—訂-----&lt; 泳 (請先閱讀背面之注意事項再填寫本頁) 本紙張尺度適用中國國家揉準(CNS ) A4規格(210X297公釐) 83. 3.10,000 五、發明説明(33 ) A7 B7 經濟部中央揉準局員工消费合作社印裝 在第八個循瓌的第二個相位(第九(P)圖),唯讀記憶 體計數器632使得唯讀記憶體630輸出零値(其對計算 g(t)没有影響)。 在第九個循環的第一個相位(第九(q)圖),多工器 626選擇上計數器622的數來爲隨機存取記憶體610定 位址。這使得隨機存取記憶體610將取樣値X(〇)輸出 至鎖存器612並將之儲存在那褢。唯讀記憶體計數器632 上計數至8。 在第九個循環的第二個相位(第九(r)圖),多工器626 選擇下計數器624的數來爲隨機存取記憶體610定位址。 回應地,隨機存取記憶體610將取樣値x(-3)輸出至加法 器614。加法器614將取樣値x(_3)與儲存於鎖存器612 的取樣値X(〇)相加,並將和X(〇)+X(-3)輸出至乘法器 64〇。同時,唯讀記憶體計數器632爲唯讀記憶體630定 位址。回應地,唯讀記憶體63〇將hi +h2値輸出至乘法 器640。乘法器640將hi +h2與χ(0)+χ(-3)相乘以產 生乘積(hi +h2 )(χ(0)+Χ(-3))。此乘積經由多工器642被 輸出至累計器644並將之儲存那兒。 上述的過程,以相類似的方式進行,以計算每一個 g(t),並如第九(s)-(z),(aa),(bb),(cc)及(dd)圖 所示。上計數器622及下計數器624均計數經過零( count through 0)。那就是,在上計數至4後.,上計數 HI- ί— Jn - - - -1 ^^1 ^ I (請先聞讀背面之注意事項再填寫本頁) 訂1 本紙張尺度適用中國國家標準(CNS &gt; A4規格(210X297公釐) 83. 3.10,000 五 、發明説明(3斗) A7 B7 器622計數至0。相同地,在下計數至0後,下計器624 計數至4。 下計器624的數是在第六個循環的第二個相位(第九 (1)圖,被儲存在鎖存器623的下計數器中。多工器626 的輸出値在第十二個循環的第二個相位(第九(X)圖)及在 第十五個循環的第一個相位上(第九(CC)圖)被儲存在鎖 存器621的上計數器中。 在每一個時間t mod L=l,一個新的取樣値 t-k -i)被寫在隨機存取記憶體610中◊例如 I-----------f 袭-- (請先閲讀背面之注$項再填寫本頁)Pi (.3), P2 (〇),..., P2 (3),..., P4 (3). However, there are only 8 unique tap coefficient values hD -h7, which are taken out as a function shown in the eighth figure. These unique coefficient values hD -h7 are assigned to any filter row and are summarized in Table 1 below: nn — J nn III (please read the precautions on the back before filling this page) -Subscribe-Table 1 Value h〇 hi h2 h3 h4 h5 he h? Tap Pi (〇) P2 (0) P3 (0) P4 (0) Pi (i) p2 (i) p3 (1) p4 (1) Tap P4 (3) P3 (3) P2 (3) Pi (3) P4 (2) P3 (2) p2 (2) Pi (2) Printed by the Peking Consumer Cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs to substitute these tap point coefficients hQ -h7 After equations (11) and (I2), the characteristics of the filter bank that has been merged between each t and ti will change, as described below: This paper scale is applicable to the Chinese National Standard (CNS) A4 specification (210X297 mm) 83.3.! 〇, 〇〇〇3 ^ 〇794 A7 ________B7 V. Description of the invention (25 *) Table of merged filter rows modified between t and ti t = l 典 t = 0 y! (1 ) = yi (l> + y4 (l) = h〇 (x (0) + x (-4)) + (h3 + h4) (χ (-1) + χ (-3)) ♦ 2h7 x (- 2) t = 2 Code t = l y2 (2) = y2 (2) * y3 (2) = h1 (x (0) * x (-4)) + (h2 + h5) (x (-1) + x (-3)) ♦ 2ht x (-2) t = 3 and t = 2 yz (3) = y2 (3) * ya (3) = (hi + hz) (x (0) + x (-3 )) + (hs * h4) (x (-1) * x (-2)) t = 4 Code t = 3 yi (4) = yi (4) * y4 (4) = (h〇 + h3) (x ( 0) * x (-3)) + (h4 * h7) (x (-1) * x (-2)) t = 5 Code yi (5) = yL (5) + y4 (5) = ha (x (1) + x (-3)) + (h3 + h4) (x (0) * x (-2)) + 2h? X (-l) t = 6 and t = 5 y2 (6) = y2 ( 6) + ya (6) = hl (x (1) + x (-3)) + (h2 + h5) (x (0) + x (-2)) 2h &amp; x (-l) t = 7 And t = 6 yi (7) = yz (7) + y3 (7) = (ht + h2) (x (1) + x (-2)) + (h5 + h6) (x (0) + x ( &quot; 1) &gt;----; ----- 威威 ------ (please read the notes on the back before filling in this book) Printed by Beigong Consumer Cooperative, Central Bureau of Standards, Ministry of Economic Affairs More valuable is that g (t) can add the output value of the merged filter row that changes the value between ti and t] (t) to the merged filter that does not change the value between tl and t The output value of the device is yi (t), y 〖(t), 83. 3.10,000 The paper size is applicable to the Chinese National Standard (CNS) Α4 specification (210X297 mm) A7 ________B7 V. Description of invention (26) Central Standard of the Ministry of Economic Affairs Bureau staff consumer cooperatives printed ..., yj-1 (t) »yj + 1 (t), yj + 2 (t), ..., y ^ J (t) were obtained. In the case of L = 4, there are only two merger filter banks. That is, a filter bank output value (t) that changes the value between t-1 and t and a filter bank output value y] (t) that does not change the value between t-1 and t. Now let ’s discuss the operation of the present invention in the ninth figures (a)-(z), (aa), (bb), (cc) and (dd) under the condition of L == 4 and N = 4. First, the first Figure 9 depicts the states of counters 622, 624, and 632, random access memory 610, and read-only memory 630. The read-only memory has 14 storage cells used to store the tap coefficients and the sum of the tap coefficients of the cross-symmetric filter. According to the changed merged filter bank shown in Table 2, the tap coefficients or the sum of tap coefficients will be grouped. This combination is separated by zero values. The random access memory 610 has n + 1 or 5 cells. It is first used to store 5 sample values x (〇), x (-1), x (-2), x (-3) and x (-4) ). For L time t-k periods, a new sampling value X (I--i) is shifted into the cell of random access memory 610 used to store the oldest sampling value. ----:-: --- ί Pack ----- ^-丨 Order ----- ί Bismuth (please read the precautions on the back before filling in this page) This paper size is suitable for Chinese national standards (CNS) Α4 specification (210X297mm) 83. 3.10,000 3 20794 A7 B7 V. Description of invention (27) The Ministry of Economic Affairs Central Standards Bureau Beigong Consumer Cooperative printed and read-only memory counter 632 was set from the initial value to read-only In the first cell of the memory 630 (for example, set to 0), the upper counter 622 is set to the first cell of the random access memory 610 (for example, set to 0) and the lower counter 624 is set Set the initial value point to the last cell of the random access memory 610 (for example, set to 4). The upper count latch 621 and the lower count latch 623 are set to initial values of 0 and 4 respectively. In the ninth (a) diagram of the first phase of the first clock pulse, the multiplexer 626 selects the count of the upper counter 622 to determine the address of the random access memory 610. The random access memory 61o therefore outputs the sample value x (0) indicated by the upper counter 622. The sample value x (〇) is stored in the latch 612. At the second phase of the first pulse clock (the ninth (b) diagram), the up counter 622 counts up to 1. The multiplexer 626 selects the number of the counter 624 to determine the address of the RAM 610. In response, the random access memory 610 outputs the sample value x (_4). The adder 614 adds the sample value x (_4) to the sample value x (0) stored in the latch 612◊The adder 614 outputs the sum x (〇) + x (_4) to the multiplier 64. Further, the read-only memory 63o outputs the h0 value addressed by the read-only memory counter 632 to the multiplier 640. The multiplication 640 multiplies h〇 with the sum x (〇) + Χ (_4) and outputs the product (eg, row filter term) hD (Χ (0) + x (-4)) to the multiplexer 642. This multiplexer outputs the product h0 (Χ (0) + x (-4)) to the accumulator 644. In the first phase 9 of the second cycle (the ninth (c) diagram), the multiplexer 626 selects the number of the upper counter 622 to determine the random access memory 610 ----.----- f Installation — (Please read the precautions on the back before filling in this page) This paper size is printed in Chinese National Standard (CNS) A4 (210X297mm) 83. 3.10,000 Printed by the Consumer Labor Cooperative of the Central Escalation Bureau of the Ministry of Economic Affairs Prostitute 07 94 A7 __ B7 V. Description of invention (2 points) Address. In response, the random access memory 610 stores the sample value x (-1) in the latch 612. Simultaneously, the down counter counts down to 3 and the read-only memory counter 632 counts up to 1. On the two phases of the second cycle of sound (the ninth (d) diagram), the multiplexer 626 selects the number of the counter 624 to locate the random access memory 610. In response, the random access memory 610 outputs the sampling value χ (-3) to the adder 614, where it is added to the sampling value X (-1) in the latch 612. Simultaneously, the read-only memory 630 outputs a combination of tap point coefficients h3 + h4 addressed by the counter 632. The sum h3 + h4 and the sum. X (-l) + X (_3) are multiplied in the multiplier 64〇 and the product (h3 + h4) (X (-l) + x (-3)) thus formed is Output to the multiplexer 642. The multiplexer 642 selects the product (h3 + h4) (χ (-1) + Χ (-3)) output from the multiplier 640 as the input value of the input accumulator 644. The accumulator 644 adds the product (h3 + h4) (χ (-1) + χ (_3)) to the product h〇 (Χ (0) + Χ (-4)) that has been stored in the n ° ° in the third On the first phase of the loop (Figure 9 (e)), the multiplexer 626 selects the number of the upper counter 622 to locate the random access memory 61. In response, the random access memory 610 stores the output of the sampling value χ (-2) in the latch 612. Simultaneously, the down counter counts down to 2 and the read-only memory counter 632 counts up to 2. On the second phase of the third cycle, the ninth (f) figure multiplexer 626 selects the number of the counter 624 to locate the random access memory 610. The random access memory 610 outputs the sample value x (-2) to the adder 614, where it is added to the sample value X (-2) in the latch 612, and at the same time, the read-only memory 63 will be Read the tapping point coefficient I of the memory 632 location address ---------- f · install ------ order ----- (please read the precautions on the back before filling this page) This paper scale is applicable to the Chinese National Standard (CNS) A4 specification (210X297 mm) 83. 3.10,000 32〇tg4 A7 ___________B7 V. Description of invention (2?) Printed by the Industrial and Consumer Cooperative Cooperative of the Zhongshang Standards Bureau of the Ministry of Economic Affairs . The tap coefficient h7 is multiplied by 2 x (-2) in the multiplier 640, and the product 2h7 x b 2) thus formed is output to the multiplexer 642. The multiplexer 642 selects the product 2h? X (_2) output by the multiplier 64〇 as the input value of the accumulator 644. The accumulator 644 adds the product 2h7 χ (-2) to the product sum (h3 + h4) (χ (-ι) + χ (-3)) + h0 (x (0) + x (-4) ) Add. Therefore, the accumulator 644 produces the output 〆yi (t). Up and down counters 622 and 624 on the first phase of the fourth cycle (the ninth (g) diagram) will be stored with the latches 621 and 623 respectively connected to them. Number of loads. Further, the read-only memory counter 632 counts down to 3. The product sum 2h7 x (-2) + (h3 + h4) (x (-i) + x (-3)) + h〇 (x (0) + X (_4)) is stored in the accumulator 644 and is Output to latch 636 and store there. The product represents the portion of g (t + i) that will not change between t and t + i. That is, the output value yl (t) of the merge filter row that does not change between t = l and t + l = 2. On the second phase of the fourth cycle (Figure 9 (h)), the read-only memory counter 632 locates the read-only memory 630. However, the value of the located address is zero, and therefore, only the zero value is output from the multiplier 640. On the first phase of the fifth cycle (Figure 9 (i)), the multiplexer 626 selects the number of the up counter 622 to address the random access memory 610. This causes the sample value X (O) to be output from the random access memory 610 and stored in the latch 612. Read-only memory counter 632 counts up to 4 〇 (please read the precautions on the back before filling in this page) This paper size is suitable for China National Standard (CNS) A4 specification (210X297mm) 83. 3.10,000 A7 ____B7 3. Description of the invention (3) On the second phase of the fifth cycle printed by the Beigong Consumer Cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs (Figure 9 (j)), the multiplexer 626 selects the number of counters 624 to be Random access memory 610 location address. This causes the sample value output from the random access memory 610 to be input to the adder 614. This adder adds the sample value X (-4) to the sample value χ (〇> stored in the latch 612, and outputs the sum χ (_4) + χ (0) to the multiplier 640. Read only The memory counter 632 locates the address in the read-only memory 630 so that the hi value is output to the multiplier 640. The multiplier 640 multiplies the hi value by the sum x (-4) + x (0), and will The product hi (x (-4) + X (0)) is output to the multiplexer 642. The multiplexer 642 outputs the product hi (X (-4) + X (0)) to the accumulator 644, where the product is Store it. On the first phase of the sixth cycle (the ninth (K) diagram), the multiplexer 626 selects the number of the up counter 622 to locate the random access memory 610. This makes the sampling value X (-l ) Is output from the random access memory and stored in the latch 612. The read-only memory counter 632 counts up to 5. Further, the down counter 624 counts down to 3. On the second phase of the sixth cycle ( (Ninth (1)), the multiplexer 626 selects the number of the counter 624 as the random access memory 610 location address. This causes the sample value x (_3) to be output from the random access memory 61〇 to the adder 614. This adder will sample the value X (-3) is added to the sample value X (-l) stored in the latch 612 and the sum x (_3) + x (_l) is output to the multiplier 640. The read-only memory counter 63 2 will be in The read-only memory 630 there is located so that the h2 + h5 value is output to the multiplier 640. The multiplier 640 multiplies the h2 + h5 value by the sum x (-3) + x (-l) and multiplies the product ( h2 + h5) (Χ (-3) + Χ (-1)) output to the multiplexer 642, the multiplexer 642 outputs the product (h2 + h5) (x (-3) + x (-l)) to the cumulative I ------; I:-^ -installed ------ ordered (please read the $ item on the back and then fill in this page) This paper size is applicable to China National Standard (CNS) A4 specification (210X297 Mm) 83. 3.10,000 V. Description of the invention (31) A7 B7 The Ministry of Economic Affairs, Central Bureau of Economic Development, Beigong Consumer Cooperative printed a meter 644, where the product is added to the product hi (X (X ( -4) +} £ (〇)). The number of the down counter 634 is stored in the latch 623 ° The first phase of the seventh cycle (the ninth (m) figure) multiplexer 626 is selected up The number of the counter 622 is used to locate the random access memory 610. This allows the random access memory 610 to store the samples stored in the latch 612 X (-2) output. The read-only memory counter 632 counts up to 6. Further, the down counter counts down to 2. The second phase of the seventh cycle (the ninth (η) diagram), multiplexing The device 626 selects the number of the counter 624 to locate the random access memory 610. This causes the random access memory 610 to output the sampling value χ (-2) to the adder 614. The adder 614 adds the sample value x (-2) to the sample value x (-2) stored in the latch 612 and outputs the · 2Χ (_2) value to the multiplier 640. At the same time, the read-only memory counter 632 positions the read-only memory 630 there, so that the read-only memory 630 outputs the h6 value to the multiplier 64〇, and the multiplier 64〇 compares the h6 value with 2 x (_2) Multiply and output the product 2h6 X (-2) to the multiplexer 642. The multiplexer 642 outputs the product 2h6 X (_2) to the accumulator 644, where it is. The product is added to the product sum stored there (h2 + h5) (x (-3) + x (-l)) + hi (X (-4) + X (0)). The sum (h2 + h5) (xi-SJ + xi-lH + hi (x (-4) + x (0)) + 2h6 x (-2) generated by the accumulator 644 is the combined filter row yl ( t) The output value. On the first phase of the eighth cycle (the ninth (〇) diagram), the upper counter 622 and the lower counter 624 will be connected to the latch 621 or 623 m. 1 ^ 1 m I (please read the precautions on the back before filling in this page)-Order &quot; This paper size is applicable to China National Standard (CNS) A4 specifications (210X297mm) 83. 3.10,000 320794 A7 B7 V. Invention description (more 2) The Ministry of Economic Affairs, Central Bureau of Accreditation and Employee Consumer Cooperatives printed and stored separately. However, the first phase of the first cycle (ninth (a) diagram) and the first phase of the fourth cycle ( (The ninth (g) diagram) is different. The number input to the down counter 624 is 3. (In the second phase of the sixth cycle, this number has been stored in the latch 623). The read-only memory counter 632 counts up to 7. The multiplexer 642 selects the sum corresponding to g (2 &gt; unchanged part) as the input value of the accumulator 644. As a result, the accumulator outputs the unchanged filter row output value yi (1 ) Is added to the changed filter bank output value h (2) just calculated. This sum corresponding to g (2) is output to the accumulator 644. As shown in Table 2, the combined filter bank The output value (2) changes again between t = 2 and t + l = 3, for example, y2 (3) is not equal to (2). However, yi (t) does not change between t = 2 and t + l = 3 Will change. Therefore, the combined filter bank output value y2 (2) will not be stored in the latch 646. Instead, the latch 646 retains the combined filter bank output value Yi (t) to determine g (3). : —.--- f.- 装 ----- ^ — 定 ----- &lt; swimming (please read the precautions on the back first and then fill out this page) Standard (CNS) A4 specification (210X297mm) 83. 3.10,000 V. Description of invention (33) A7 B7 The Ministry of Economic Affairs Central Bureau of Labor and Industry Employees Consumer Cooperative printed on the second phase of the eighth cycle (ninth) (P) graph), the read-only memory counter 632 causes the read-only memory 630 to output a zero value (which has no effect on the calculation of g (t)). At the first phase of the ninth cycle (ninth (q) graph ), The multiplexer 626 selects the number of the counter 622 Addressing the random access memory 610. This causes the random access memory 610 to output the sample value X (O) to the latch 612 and store it there. The read-only memory counter 632 counts up to 8. In the second phase of the ninth cycle (ninth (r) diagram), the multiplexer 626 selects the number of the down counter 624 to locate the random access memory 610. In response, the random access memory 610 outputs the sample value x (-3) to the adder 614. The adder 614 adds the sample value x (_3) to the sample value X (〇) stored in the latch 612, and outputs the sum X (〇) + X (-3) to the multiplier 64. At the same time, the read-only memory counter 632 addresses the read-only memory 630. In response, the read-only memory 63o outputs hi + h2 to the multiplier 640. The multiplier 640 multiplies hi + h2 and χ (0) + χ (-3) to produce a product (hi + h2) (χ (0) + Χ (-3)). This product is output to the accumulator 644 via the multiplexer 642 and stored there. The above process is carried out in a similar way to calculate each g (t), as shown in the ninth (s)-(z), (aa), (bb), (cc) and (dd) diagrams . Both the upper counter 622 and the lower counter 624 count through zero (count through 0). That is, after counting up to 4., counting up HI- ί — Jn----1 ^^ 1 ^ I (please read the precautions on the back and then fill out this page) Order 1 This paper size is suitable for Chinese countries Standard (CNS &gt; A4 specification (210X297mm) 83. 3.10,000 V. Description of the invention (3 buckets) A7 B7 device 622 counts to 0. Similarly, after counting down to 0, the counter 624 counts to 4. The number of the counter 624 is the second phase in the sixth cycle (the ninth (1) diagram, which is stored in the down counter of the latch 623. The output value of the multiplexer 626 is the first in the twelfth cycle The two phases (ninth (X) graph) and the first phase in the fifteenth cycle (ninth (CC) graph) are stored in the up counter of the latch 621. At each time t mod L = l, a new sampling value tk -i) is written in the random access memory 610 ◊for example I ----------- f attack-- (please read the note $ item on the back first (Fill in this page again)

L 在t=5,x(l)被位移入位址格 (addressed cell) 4。當 t-k 一個新的取樣値x( -i)被寫在隨機存取記憶體 訂At t = 5, x (l) is shifted into the addressed cell (addressed cell) 4. When t-k a new sample value x (-i) is written in random access memory

L 610中,一個適當的控制訊號在控制線617 (低訊號)上被 傳送。這使得資料能夠被寫進隨機存取記憶體610中,並使 得資料能夠透過三態緩衝器(tristate buffer) 616傳送 經濟部中央標準局貝工消費合作社印製 至匯流排(bus) 618。因此,一個輸入取樣値 t-k •i)在匯流排上被輸出且被寫入隨機存取記In L 610, an appropriate control signal is transmitted on control line 617 (low signal). This allows data to be written into random access memory 610, and enables data to be transmitted through tristate buffer 616 to be printed on bus 618 by the Beigong Consumer Cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs. Therefore, an input sample value t-k • i) is output on the bus and written to the random access memory

L 本紙張尺度適用中國國家揉準(CNS ) A4規格(210X297公釐) 83. 3.10,000 ^0794五、發明説明(35) A7 B7 經濟部中央橾準局負工消費合作社印裝 t-k 憶體610中。當輸入取樣値x( - -i)被寫入後, 一 L - —個商訊號(high signal)在控制線61*7上被輸送,此 時,可以從隨機存取記憶體610再讀取取樣値,而且透 過三態緩衝器616使其不傳輸資料。 在後繼的過程中,t=5所決定的已改變的且已合併濾 波器排輸出値y'2 (t)被儲在鎖存器646之中,因爲此 輸出値y2 (t)在t = 5及t=6之間不會改變。另一方面, 已合併濾波器排輸出値yl (t)在t=5及t=6之間會改 變其値,所以可棄之。 在計算每一個g(t)時,增頻器6〇〇以加法器614 η 爲準備(proliminarily)要相加的取樣値執行-個 2 相加。然而,這可用電路62〇,610,612及614來取代 而得到改善,這樣的一個電路可將取樣輸入値適當的和儲 存起來,並且在適當的時間將它們輸出至乘法器64〇。此 (請先閲讀背面之注意事項再填寫本頁) 本紙張尺度適用中鬮國家橾準(CNS }八4说格(21〇χ297公釐) m _〇〇 A7 B7五、發明説明(36 ) 經濟部中央標準局貝工消费合作社印製 外,乘法器64〇在將取樣値與分接點係數或分接點係數的 η 和相乘時,至多執行-+ 1個相乘。累計器644爲相 2 η 加排濾波器項需要執行——個相加,以便形成濾波器排 2 輸出値y] (t)°進一步地,累計器644爲相加濾波器排 κ L 輸出値yi (t)需要執行-個相加以便形成g (t)。 2 大不相同地,濾波器200 (第三圖)至少使用η個相乘 及幾個相加來計算每一個g(t)° 增頻器600在n=4且L=4的情況下’已被描述過 了。然而,一個增頻器600可被用在任何η及L的情 況下。隨機存取記憶體61〇及唯讀記憶630必須根據n 的不同而調整。此外,儲存在唯讀記憶體630中的分接 點係數必須根據η及L重新計算。進一步地,計數器 622,624及632,多工器626及鎖存器621及623的 順序(sequencing)必須被適當地修正。 當L=2時,僅存在著一個合併濾波器排。此濾波器 排在每一個t-l及t之間改變。因此,鎖存器646及 工器642可被移去。取而代之的,累計器與相乘器64〇 的輸出値直接連接。 (請先閲讀背面之注意事項再填寫本頁) 装· 本紙張尺度適用中國國家橾準(CMS〉A4規格(210X297公釐) 83. 3. !0,〇〇〇 A7 _B7五、發明説明(37) 經濟部中央橾準局員工消費合作社印裂 對L&gt;4,多於一個的合併濾波器排輸出値y〗(t)在 t-ι及t之間保持不變。因此,鎖存器646可被一個多 工器及一個用來儲存每一個不變的濾波器排輸出値yi (t) 的鎖存器所取代。多工器被連接至累計器644的輸出値 並且交替地將在累計器644中產生的値儲存在每一個鎖存 器中。同樣地,多工器642可以被適當地修正爲逐一地選 擇,以循環方式儲存在這些鎖存器中的每一個不會改變的 濾波器排輸出値(t),並與在累計器644中的已計算的 且已改變的濾波器排輸出値(t)相累計以決定每一個 g(t)的値。第十圖是表示根據本項發明使用在一個視訊應 用的場合中的第二個實施例的一個增頻器700。增頻器700 在某些方面與第五圖的傳統的有限脈衝反應濾波器400相 類似,增頻器700有一個模組核心710及一個有加法器 791的二進制相加樹(a binary adder tree) 790差別之 處是,以交叉對稱排濾波器項合併的方式以合成濾波器排, 來取代以合併輸入取樣値的位元面(bit planes)方式( 使用第q個,第(q+1)個位元鄰近對來執行計算。因此, 基於前述有關有限脈衝反應濾波器400的技巧的處理需 ---------f -裝-----1 訂------i h (請先閲讀背面之注意事項再填寫本頁) 本紙張尺度適用中國國家標準(CNS &gt; A4規格(21〇X297公釐) 83. 3.10,000 ^0794五、發明説明(3牙) A7 B7 求(所需電路處理時間)主要是依賴輸入取樣値的位元b 的數量,增頻器700的處理需求依賴分接點係數的數量η。 第十一圖表示模組核心710的一個實施例。模組核心 •710有一個將已輸入的取樣値χι t-j? i)選擇性L This paper scale is applicable to China National Standard (CNS) A4 (210X297mm) 83. 3.10,000 ^ 0794 V. Description of the invention (35) A7 B7 Central Government Bureau of Economic Affairs, Ministry of Economic Affairs, Unemployed Consumer Cooperatives printed tk memory 610. When the input sample value x (--i) is written, an L-a high signal is sent on the control line 61 * 7, at this time, it can be read again from the random access memory 610 Sampling value, and through the three-state buffer 616 so that it does not transmit data. In the subsequent process, the changed and merged filter bank output value y'2 (t) determined by t = 5 is stored in the latch 646 because the output value y2 (t) is at t = No change between 5 and t = 6. On the other hand, the output value yl (t) of the combined filter bank will change its value between t = 5 and t = 6, so it can be discarded. In the calculation of each g (t), the frequency multiplier 600 performs a 2-addition with the adder 614 η prepared to proliminarily sample the value to be added. However, this can be improved by replacing it with circuits 620, 610, 612, and 614. Such a circuit can properly store and store the sample input values and output them to the multiplier 64 at the appropriate time. This (please read the precautions on the back and then fill out this page) This paper scale is applicable to the Chinese National Standard (CNS) 八四 说 格 (21〇χ297mm) m _〇〇A7 B7 V. Description of the invention (36) Printed by the Beigong Consumer Cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs, when multiplying the sample value by the tap coefficient or the η sum of the tap coefficient, the multiplier 64 will perform at most-+ 1 multiplication. Accumulator 644 Adding the filter term for the addition of 2 η needs to be performed — an addition to form the output value of the filter row 2] (t) ° Further, the accumulator 644 outputs the output value of the addition filter row κ L yi (t ) Needs to perform an addition to form g (t). 2 Very differently, the filter 200 (third figure) uses at least η multiplications and several additions to calculate each g (t) ° frequency increase 600 has been described in the case of n = 4 and L = 4. However, a frequency multiplier 600 can be used in any case of n and L. Random access memory 61〇 and read-only memory 630 must be adjusted according to the difference of n. In addition, the tap coefficient stored in the read-only memory 630 must be recalculated according to η and L. Further, The sequence of counters 622, 624 and 632, multiplexer 626 and latches 621 and 623 must be properly modified. When L = 2, there is only one merge filter bank. A tl and t change. Therefore, the latch 646 and the worker 642 can be removed. Instead, the accumulator is directly connected to the output value of the multiplier 64. (Please read the notes on the back before filling in This page is loaded · This paper scale is applicable to the Chinese National Standard (CMS> A4 size (210X297 mm) 83. 3.! 0, 〇〇〇A7 _B7 V. Invention description (37) Employee consumption of the Central Standards Bureau of the Ministry of Economic Affairs Cooperative society splits L> 4, the output value of more than one merge filter row (t) remains unchanged between t-ι and t. Therefore, the latch 646 can be used by one multiplexer and one To store the output value yi (t) of each unchanged filter bank. The multiplexer is connected to the output value of the accumulator 644 and alternately stores the value generated in the accumulator 644 in each A latch. Similarly, the multiplexer 642 can be appropriately modified to select one by one, The output value (t) of the filter bank stored in each of these latches in a cyclic manner is unchanged, and is accumulated with the calculated and changed output value of the filter bank (t) in the accumulator 644 To determine the value of each g (t). The tenth figure shows a frequency multiplier 700 of the second embodiment used in a video application according to the present invention. The frequency multiplier 700 differs from The conventional finite impulse response filter 400 in the fifth diagram is similar. The frequency multiplier 700 has a module core 710 and a binary adder tree 790 with an adder 791. The difference is that The symmetric row filter term combination method uses a synthesis filter row to replace the bit planes method of combining input sampling values (using the qth, (q + 1) th bit neighboring pairs to perform calculations . Therefore, the processing based on the aforementioned technique of the finite impulse response filter 400 requires --------- f -installed ----- 1 order ------ ih (please read the notes on the back first (Fill in this page again) This paper scale is applicable to the Chinese National Standard (CNS &gt; A4 specification (21〇X297mm) 83. 3.10,000 ^ 0794 V. Invention description (3 teeth) A7 B7 request (required circuit processing time) It mainly depends on the number of bits b of the input sampling value, and the processing demand of the multiplier 700 depends on the number of tap coefficients η. The eleventh figure shows an embodiment of the module core 710. The module core 710 has one Will the input sampling value χι tj? I) Selectivity

L 地相加在一起,以形成取樣値的和的先前處理級(Preprocessing stage) ° 此先前處理 級有一 個位移暫存器 (shift register) 722。該暫存器有箸n+1個暫存器, 縱排地連接在一起(connected together in tandem)。 此先前一處理級720也有U+l個多工器724-0 , 724-1, (請先閲讀背面之注意事項再填寫本頁) 裝. 訂. -)。每第U個多 經濟部中央橾準局貝工消費合作社印製 η » 724-η » ...,724-U (其中 U= 工器724-U (其中〇SuSU)。除了多工器724-u,將接收 t&quot; t- I 取樣値X ( -n+u-i+1) , X( L L 」 — L _ -n+u-i) 而此取樣値被儲存在暫存器722-(n-u-l)及722-n-u中。 例如,多工器724-0分別自儲存在暫存器722-(n-l)及 t- s 722-n中接收到取樣値x(The L grounds are added together to form the preprocessing stage of the sum of the sampling values. This preprocessing stage has a shift register 722. The scratchpad has n + 1 scratchpads, which are connected together in tandem. This previous processing stage 720 also has U + l multiplexers 724-0, 724-1, (please read the precautions on the back before filling out this page) to install. Order.-). Every Uth Multi-Ministry of Economic Affairs Central Bureau of Industry and Commerce Beigong Consumer Cooperative Printed η »724-η», 724-U (where U = worker 724-U (where 〇SuSU). Except for multiplexer 724 -u, will receive t &quot; t- I sample value X (-n + u-i + 1), X (LL ”— L _ -n + ui) and this sample value is stored in the register 722- (nul ) And 722-nu. For example, the multiplexer 724-0 receives sampling values x () from the registers 722- (nl) and t-s 722-n, respectively.

L 83. 3.10,000 本紙張尺度逋用中國國家標準(CNS ) A4規格(210X297公釐)L 83. 3.10,000 The size of this paper adopts the Chinese National Standard (CNS) A4 specification (210X297mm)

X I t-1X I t-1

L 器 722- η A7 B7 多工器724_u將接收到儲存於暫存 的値及零値當作輸入値。多工器 724-0 ··· ’ 724_u在—個外部產生訊號ct2的控制下選擇了輸 入値中的一個。 此先目丨〗~處理級72〇也有U個加法器726-0,726-1 &quot;· , 726_u,·_·,。每一第u個加法器(其中 HUSU-I) 726_u將儲存在第u個暫存器722_u中的取樣 値X( t-The L device 722- η A7 B7 multiplexer 724_u takes the value and zero value stored in the temporary storage as input values. The multiplexer 724-0 · ’724_u selected one of the input values under the control of an externally generated signal ct2. This preface 丨〗 ~ The processing stage 72〇 also has U adders 726-0, 726-1 &quot; ·, 726_u, · _ · ,. Each u-th adder (of which HUSU-I) 726_u will store the sample stored in the u-th temporary register 722_u-X (t-

L •U-i)相加至被第u個多工器所選擇的 取樣値。這些取樣値的和適合與相對應於合併濾波器排的 分接點係數値相乘。此取樣値及取樣値的和自加法器726 -0,. · .,726-U-1及多工器724-U被平行地輸出至一個 相乘級730。 ---:---:---1 '赛-----1-5 (請先閲该背面之注意事項再填寫本頁) 經濟部中央橾準局員工消费合作社印製 本紙張尺度適用中國國家揉準(CNS ) A4規格(210X297公釐) 83. 3.10,000 A7 B7五、發明説明(4〇) 經濟部中央標準局貝工消費合作社印策 相乘級73〇有U+1個分接點係數選擇電路器( selector circuits) 735垂0,735-1 ,…,735-u ,…, 735_U,其被縱排地(j_n tandem)的連接在一起。每一個 選擇電路735-0,...,735-U相對應於每一個合併濾波器 排的一個特定的合併排濾波器項選擇電路735-0 , 735_U在每一個時間t平行地輸出一個分接點係數値。 每一個選擇電路735-u包括一個與一個位移暫存器734-u 相連接的多工器732-u。 例如,選擇電路735-1包括一個與位移暫存器734-1 相連接的多工器732-1。多工器732-U自其本身的位移 暫存器734_u輸出的分接點係數値,或者自前一個選擇電 路73〇-u-l的位移暫存器734-u-l輸出的分接點係數値, 兩者之中,依一個外部控制訊號Ctl,選擇其一。(多工器 732-0是例外,其自本身位移暫存器734-0輸出的分接 點係數値,或者自輸入在線736上的分接點係數値兩者之 中,選擇其一。) 位移暫存器734-u儲存了 L個分接點係數値,亦即, 在每一個合併濾波器排下的一個分接點係數値。這些分接-點係數値以相同的次序被儲存在每一個位移暫存器734-U 中。此分接點係數値被接回(fed back)其、相對應的多工 器。因此,每一個位移暫存器734-U循環地重覆在時間t 的每一個L時間週期(time periods),自其中輸出分接 點係數。 相乘級730有U+1個乘法器740-0,740-1,..., 740-U,...,740-U。這些乘法器740-u可能會是任何一 (請先閲讀背面之注意事項再填寫本頁) •裝· 本紙張尺度適用中國國家揉準(CNS ) A4規格(210X297公釐) 83. 3. 10,000 五、發明説明(41 ) A7 B7 經濟部中央標準局另工消費合作杜印製 種乘法器,包括最佳Booth's運算法則乘法器(optimized Booth7s algorithm multipliers) 〇 每一第 u 個乘法器 74〇-u接收到一個相對應的分接點係數選擇器735_u的輸 出値。(特別地,分接點係數値自位移暫存器級734-u位 移出來)。每一個乘法器740-0,740-1,...,740-U-1將 分接點係數値與先前一處理階級720的相對應的加法器 726-0,726-1,·..,726-U-1的输出値相乘。乘法器 74〇-U將其接收到的分接點係數値與多工器724_u的輸 出値相乘。 每一個乘法器740-U輸出一個相對應的合併排濾波 器項yxD ,yxi ,. . .,yxu . . ,,yxv,這些合併排濾 波器項 yxD ,yxj_ ,. . ·,yxu . . .,yxu 每一個均被二 進制相加樹79〇 (710)的一個相對應的二進制加法器791 (第十圖)所接收入。此二進制相加樹790在任何一個時 間t將合併排滤波器項yx〇 ,,· . .,yxu . · ·,yxu 予以相加。此二進制相加樹790輸出y〗(t)。 鎖存器次電路(sub-circuit) 760包括J個個別鎖I~ b π 存器(其中J=-)。爲了要計算每一個g(t),首 2 先,將已改變的合併排濾波器輸出値y] (t)儲存在被多 -I- —J I - -- —-1 -装-- (請先閲讀背面之注意事項再填寫本頁) 訂 4- 本紙張尺度逋用中國國家標準(CNS ) A4規格(210X297公釐) 83.3. !0,〇〇〇 A7 _B7五、發明説明(令2) 經濟部中央標準局男工消费合作社印装 工器76〇選擇的鎖存器次電路77〇的第j個鎖存器之 中。然後控制線755被用來歸零(clear)累計器750。 第二步,多工器78〇選擇每一個自1至J的第j 個鎖存器的輸出値,(其儲存了第j個合併排濾波器输出 値y/(t))。被選擇的第j個鎖存器輸出値被輸入至累 計器75〇的一個輸入値。藉由計算所有的合併排濾波器 輸出値y/(t),累計器75〇產生增頻取樣輸出値(the interpolated sample output) g(t) 〇 以下將詳盡地探討第十圖及第十一圖增頻器700的 運作。假設增頻器700是一個L=4相位,n=4分接點係 數的增頻器,其以d位元分接點係數p來增頻b位元 t-jj 取樣値X( - -i)。在這種惰形下,u=2。首先ctl 一 L 一 被設定爲零。因此,乘法器級的多工器732-0選擇輸入 在輸入線736上的分接點係數。另外的多工器732-1及 732-2選擇輸出自先前選擇電路735-0或735-1的分接 點係數値。位移暫存器734_〇,734-1及734-2藉由輸 入線736,首先載入它們的分接點係數値。表三摘要了載 I----------&lt; 裝------—訂-----^ i (請先閱讀背面之注意事項再填寫本頁) 本紙張尺度適用中國國家標準(CNS ) A4規格(210X297公釐) 83. 3.10,000 32〇794 at — B7 明發 入每一個位移哲存器734-0,734-1及 734_2分接點係數値的情形: 表三 位移暫存器 分接點係數値(依輸出値的次序··從左至右) 734-0 734-1 734-2 h〇 f hi 9 hj^ +h2 ,h〇 +1¾ h3 +¾ . , h2 +h5 , h5 , 1¾ +h7 2h7 ,2h&amp; ,0,0 一裝-- (請先閱讀背面之注意事項再填寫本頁) 經濟部中央標準局員工消費合作杜印製 然後,ctl被設定爲以致於多工器732_〇, 7324及 732-2選擇了自相對應的位移暫存器734_〇 , 7344及 734~2輸出的分接點係馼値。對時間t的每-個L時間 迎期,重覆執行下列步驟。首先,下一個(next)取樣値 x( - -i)被位移入先前一處理級的位移暫存器 — L 」 7 22。訊號Ct2被設爲1,以致於多工器724-u選擇了 儲存在暫存器724_(n-u)中的取樣値。也就是,多工器 724-0選擇了儲存在盥存器722_4中的取樣値,多工器 724_1選擇了儲存在722_3中的取樣値且多工器724-2 選擇了儲存在722_2中的取樣値。結采,加法器726_〇 將儲存在暫存器722-0中的及72 2-4中的取樣値相加, 亦即,Χ(0)+Χ(_4)。這個和被輸出至乘法器740-0。相似 訂 本紙張尺度適用中國國家標準(CNS ) A4規格(21〇X297公釐) 83. 3. 10,000 A7L • U-i) are added to the sampling value selected by the u-th multiplexer. The sum of these sampling values is suitable for multiplying the tap point coefficient values corresponding to the combined filter bank. The sampling value and the sum of the sampling value are output from the adder 726-0,..., 726-U-1 and the multiplexer 724-U in parallel to a multiplying stage 730. --- : ---: --- 1 '赛 ----- 1-5 (please read the precautions on the back before filling this page). The paper size of the printed paper is printed by the Employee Consumer Cooperative of the Central Bureau of Economic Affairs of the Ministry of Economic Affairs. Applicable to the Chinese National Standard (CNS) A4 specification (210X297mm) 83. 3.10,000 A7 B7 V. Description of invention (4〇) Multiplied by the Central Bureau of Standards of the Ministry of Economic Affairs Beigong Consumer Cooperatives 73. U + 1 A number of selector point coefficient selector circuits (selector circuits) 735 is connected to 0,735-1, ..., 735-u, ..., 735_U, which are connected together in tandem. Each selection circuit 735-0, ..., 735-U corresponds to a specific merge row filter item selection circuit 735-0, 735_U corresponding to each merge filter row and outputs a minute in parallel at each time t Contact coefficient value. Each selection circuit 735-u includes a multiplexer 732-u connected to a shift register 734-u. For example, the selection circuit 735-1 includes a multiplexer 732-1 connected to the shift register 734-1. Multiplexer 732-U tap point coefficient value output from its own displacement register 734_u, or tap point coefficient value output from the previous selection circuit 73〇-ul's displacement register 734-ul, both Among them, choose one according to an external control signal Ctl. (The multiplexer 732-0 is an exception. It selects one of the tap coefficient values output from its own shift register 734-0 or the tap coefficient values input on line 736.) The shift register 734-u stores L tap coefficient values, that is, one tap coefficient value under each merge filter row. These tap-point coefficient values are stored in each shift register 734-U in the same order. The tap point coefficient value is fed back to its corresponding multiplexer. Therefore, each shift register 734-U cyclically repeats every L time period at time t, from which tap coefficients are output. The multiplication stage 730 has U + 1 multipliers 740-0, 740-1, ..., 740-U, ..., 740-U. These multipliers 740-u may be any one (please read the precautions on the back before filling in this page) • Installed · This paper size is suitable for China National Standard (CNS) A4 specification (210X297mm) 83. 3. 10,000 Fifth, the invention description (41) A7 B7, the Central Standards Bureau of the Ministry of Economic Affairs, the Ministry of Economy, Labor and Consumer Cooperation, the Duyin multipliers, including the optimal Booth's algorithm multipliers (optimized Booth7s algorithm multipliers). Each uth multiplier 74. u receives the output value of a corresponding tap point coefficient selector 735_u. (In particular, the tap point coefficient value is shifted out from the shift register stage 734-u bit). Each of the multipliers 740-0, 740-1, ..., 740-U-1 adds the tap point coefficient values to the adders 726-0, 726-1, ... of the previous processing class 720. , The output value of 726-U-1 is multiplied. The multiplier 74〇-U multiplies the tap point coefficient value it receives by the output value of the multiplexer 724_u. Each multiplier 740-U outputs a corresponding merged row filter term yxD, yxi,..., Yxu..., Yxv, these merged row filter terms yxD, yxj_,..., Yxu... , Yxu are each received by a corresponding binary adder 791 (Figure 10) of the binary addition tree 79 (710). This binary addition tree 790 adds the combined row filter terms yx〇, yxu, yxu, yxu at any time t. This binary addition tree 790 outputs y (t). The latch sub-circuit 760 includes J individual latch I ~ b π registers (where J =-). In order to calculate each g (t), the first 2 first, store the changed merge filter output value] (t) in the multi-I- —JI--—-1 -installed-- (please Read the precautions on the back first and then fill out this page) Order 4- This paper scale uses the Chinese National Standard (CNS) A4 specification (210X297 mm) 83.3.! 0, 〇〇〇A7 _B7 V. Description of the invention (Order 2) Among the j-th latches of the latch sub-circuit 77〇 selected by the printing tool 760 of the Male Workers ’Consumer Cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs. The control line 755 is then used to clear the accumulator 750. In the second step, the multiplexer 78 selects the output value of each jth latch from 1 to J, (which stores the output value y / (t) of the jth merge bank filter). The selected jth latch output value is input to an input value of the accumulator 75〇. By calculating the output value y / (t) of all the combined row filters, the accumulator 75 generates the up-sampled output value (the interpolated sample output) g (t). The tenth and eleventh figures will be discussed in detail below The operation of the frequency multiplier 700 is shown. Assuming that the frequency multiplier 700 is an L = 4 phase and n = 4 tap coefficient multiplier, it is multiplied by b bit t-jj sampling value X (--i ). In this inert shape, u = 2. First, ctl_L_ is set to zero. Therefore, the multiplexer stage multiplexer 732-0 selects the tap coefficient input on the input line 736. The other multiplexers 732-1 and 732-2 select the tap coefficient values output from the previous selection circuit 735-0 or 735-1. The displacement registers 734_〇, 734-1 and 734-2 are first loaded with their tap point coefficient values via the input line 736. Table 3 summarizes the contents of I ---------- &lt; outfit -------- order ----- ^ i (please read the precautions on the back before filling this page) Applicable to China National Standard (CNS) A4 specification (210X297mm) 83. 3.10,000 32〇794 at — B7 Mingfa into each displacement memory 734-0, 734-1 and 734_2 branch point coefficient value : Table 3 Displacement register tap coefficient value (in order of output value ·· from left to right) 734-0 734-1 734-2 h〇f hi 9 hj ^ + h2, h〇 + 1¾ h3 + ¾., H2 + h5, h5, 1¾ + h7 2h7, 2h &amp;, 0,0 Pack-(please read the notes on the back before filling this page) Employee consumption cooperation of the Central Bureau of Standards of the Ministry of Economic Affairs is printed, then, The ctl is set so that the multiplexers 732_〇, 7324, and 732-2 select the tap point system output from the corresponding displacement registers 734_〇, 7344, and 734 ~ 2. For every L-time welcome period at time t, repeat the following steps. First, the next (next) sample value x (--i) is shifted into the shift register of the previous processing stage — L ″ 7 22. The signal Ct2 is set to 1, so that the multiplexer 724-u selects the sampling value stored in the register 724_ (n-u). That is, the multiplexer 724-0 selects the sampling value stored in the toilet 722_4, the multiplexer 724_1 selects the sampling value stored in 722_3 and the multiplexer 724-2 selects the sampling stored in 722_2 value. As a result, the adder 726_〇 adds the sample values stored in the register 722-0 and 72 2-4, that is, Χ (0) + Χ (_4). This sum is output to the multiplier 740-0. Similar to the size of this paper, the Chinese National Standard (CNS) A4 specification (21〇X297mm) is applicable 83. 3. 10,000 A7

五、發明説明(Wf) 經濟部中央揉準局貝工消费合作社印製 地’加法器726_1將儲存在暫存器722_1及722_3的 取樣値的和,即,Χ(-1)+χ(-3),輸出至乘法器74〇_1〇多 工器724_2將儲存在暫存器722_2的取樣値輸出至乘法 器 740-2,即,χ(-2)。 選擇器73〇-0將在那裏的第—個分接點係數値,例 如,hD ,輸出至乘法器74〇_〇。乘法器74〇_〇將此已修 正的係數與自加法器726-0輸出的取樣値和相乘。此乘 積如同第一個排濾波器項yxQ地被輸出。相同地,選擇 器73〇-1將儲存在那褢的第一個分接點係數値,例如h3 +h4 ,輸出至乘法器74〇-1。乘法器·740-1將此已修正 的係數與自加法器726_1輸出的取樣値的和相乘。此乘 稹被視爲第二個排濾波器項yXl般地輸出。選擇器730 _2將儲存在那衷的第一個分接點係數値,例如,2h7輸 出至乘法器74〇~2。乘法器74〇_2將此已修正的係數與 自多工器724_2輸出的取樣値相乘。此乘積被視同第三 個排濾波器項yx2般地輸出。排濾波器項yX() , yXl 及yx2被平行地輸出至二進制相加樹790 〇V. Description of Invention (Wf) Printed by the Ministry of Economic Affairs Central Bureau of Accuracy and the Beigong Consumer Cooperatives' adder 726_1 will be the sum of the sampling values stored in the registers 722_1 and 722_3, ie, Χ (-1) + χ (- 3), output to the multiplier 74. The multiplexer 724_2 outputs the sample value stored in the register 722_2 to the multiplier 740-2, that is, χ (-2). The selector 73〇-0 outputs the coefficient value of the first tap point there, for example, hD, to the multiplier 74〇_〇. The multiplier 74〇_〇 multiplies the corrected coefficient by the sum of the sample values output from the adder 726-0. This product is output as the first row filter term yxQ. Similarly, the selector 73〇-1 outputs the coefficient value of the first tap point stored in that loop, for example, h3 + h4, to the multiplier 74〇-1. The multiplier 740-1 multiplies the corrected coefficient by the sum of the sample values output from the adder 726_1. This multiplication is regarded as the output of the second row filter term yXl. The selector 730_2 will store the coefficient value of the first tap point stored in that center, for example, 2h7 and output it to the multipliers 74〇 ~ 2. The multiplier 74〇_2 multiplies the corrected coefficient by the sample value output from the multiplexer 724_2. This product is output like the third row filter term yx2. The row filter terms yX (), yXl, and yx2 are output in parallel to the binary addition tree 790.

L 在下一個循環至第k——J個循環,多工器724-0 , 724_1及724_2選擇相同的取樣値,而且加法器726_〇 及726-1輸出相同取樣値的和。選擇器735-0 , 735-1 及735-2將相應於那個循環的分接點係數値輸出。亦即, ----.-----^ -裝-----^--訂 (請先閲讀背面之注意事項再填寫本頁) 本紙張尺度逋用中國國家揉準(CNS &gt; A4規格(210X297公釐) 83.3.10,000 3807㈧ A7 B7 五、發明説明(4左) 經濟部中央橾準局tec工消费合作社印製 在第二個循環,選擇器735-0輸出hi ,選擇器735-1 輸出h2 +h5 ,以及選擇器735_2輸出2h&amp; 。乘法器 740-0,740-1,及740-2將相對應來自多工器724-2取 樣値或來自加法器724_〇及724的取樣和相乘,以產 生個別的排濾波器項 yx〇 , yxi 及 yx2 0 L 從循環 -+1開始,控制訊號Ct2被改變成 —2 」 爲〇。這使得每一多工器724-u (除了多工器724-U)去 選擇儲存在暫存器722-n-u-l的取樣値做爲加法器726-U 的輸入値。亦即,多工器724_〇選擇了儲存在722-3暫 存器中的取樣値及多工器724-1選擇了儲存在暫在器722 _2中的取樣値。多工器724_2選擇零値。因此加法器 726_0將儲存在暫存器722_0及722-3之中的取樣和輸 出,例如,χ(0)+χ(-3)。加法器726-1將儲存在暫存器 722-1及722-2中的取樣値的和輸出,即χ(-ι)+χ(-2) 〇 L 在第(-J +1)個循環,選擇器735-u將儲存在 L 那裏的第(-+1)個分接點係數値輸出。例如,在 —2 - 第三個循環,選擇器735-〇輸出係數hi +h2 ,選擇器 735-1輸出係數h5 +h6及選擇器735_2輸出零。乘法 (請先閲讀背面之注意事項再填寫本頁) 本紙張尺度適用中國國家標準(CNS ) A4規格(210X297公釐) 83. 3. 10,000 五、發明説明(与έ ) A7 B7 經濟部中央揉準局貞工消費合作社印製 器740-0接著將係數hi +h2與和X(0)+X(_3)相乘。 乘法器740-1將係數h5 +h6與和X(-l)+X(_2)相乘, 及乘法器將係數零與零相乘。這些乘積被分別當作排濾波 器項yx〇 ,yh及yx2輸出至二進制相加樹且被 累計,以決定yj (t)的値。 自下一個循環至循環L的每一個循環,多工器 724_0 , 724-1及724-2選擇了相同的取樣値且加法器 726-0及726_1輸出了相同的取樣値的和。選擇器735-0 , 735_1及735_2輸出了相對於那個循瓌的分接點係數 値。亦即,在第四個循環,選擇器735-0輸出hD +h3 , 及選擇器735-1輸出h4 +h7 ,選擇器735-2輸出零。 乘法器740-0,及740-1與自加法器724-0或724-1輸 出的相對應之取樣値的和相乘,以產生個別的已修正了的 排濾波器項yxQ ,yxi (乘法器74〇_2將係數零與自多 工器724_2輸出的零値相乘,以產生合併排濾波器項yX2 =〇。)這些合併排濾波器項yxQ ,yXl ,及yX2被二進 制相加樹79〇所累計以決定yj (t)的値。.最後,在累 計了合併排濾波器輸出値yj (t)的所有J個後,累計 器750輸出g(t)。 上述過程接著重覆著進行,亦即,下一個取樣値 t-1 x( |^— J _i + l)(被稱爲,x(l))被位移入先前—處 (請先聞讀背面之注$項再填寫本頁) .裝· 訂_ 本紙張尺度適用中國國家檬準(CNS ) Α4规格(210X297公釐) 83. 3.10,000 A7 B7 經濟部中央橾準局貝工消費合作社印製 五、發明説明(47) 理級710的位移暫存器722 ,且訊號ct2被設定爲1, 等等。開始時,位移暫存器722可能大部分是空的。因 此* g(t)的初始被決定値或許不太有用。相反的,位移 暫存器722可以被預先載入近似的取樣資料,以使得 g(t)所有的決定値都是可用的。 增頻器700較之FIR濾波器400 (第五圖)需要非 常低的處理需求。增頻器700的二進制相加樹790需要 η -個加法器,來先前處理輸入的取樣値及少於(近似 2 η 地)-個加法器用在二進制相加樹中。相乘級730需 2 η 要-+ 1個乘法器,或相等於Booth乘法器的量被 2 n b 使用的話,需要(-+ 1) ·-個加法器。因此, 2 2 根據本發明的增頻器可在近似((ϋ+1) .-^-+ η) 2 2 個加法器的情況下來執行。相反地,有限脈衝反應(FIR) b 濾波器400的二進制相加樹43〇需要大約(n+1) ·- 2 個加法器。因此傳統的有限脈衝反應濾波器40(/需要下 列更多的加法器: 本紙張尺度適用中國國家標準(CNS ) A4規格(210x297公釐) 83. 3.10,000 ^—^1, -.....I n 1 - —^ϋ In I (請先閲讀背面之注意事項再填寫本頁) 訂 A7 B7 五、發明説明(斗?) bn b n b (n+1) · - - ((- + 1) . - + n) = -- (- -2) 2 2 2 2 2 —般而言,每一個視訊取樣値均多於4個位元,亦 即,b&gt;4 °因此,使用增頻器700可達到節省的效杲。 簡短地説,本發明藉由組合交叉對稱排濾波器項, 揭露一個具有合併濾波器排的增頻器。以這種濾波器組的 合併方式,可減少處理過程的需求(亦即,電路的數量, 電路所佔據的空間,及處理時間)而不會阻礙增頻器的 正確性或效率。 惟,上述本發明所揭露之圖式説明,僅爲本發明之實 施例,大凡熟悉本案技藝之人仕,其所依本案精神範疇, 所做之等效變化或修飾之實施例,皆應涵蓋在以下本案之 申請專利範圍内。 ^^1' —^1 ^^1- ^^1 —^1 ·_ϋ— ^ (請先閎讀背面之注意事項再填寫本頁) 訂_ 經濟部中央標準局貝工消費合作社印製 83. 3.10,000 本紙張尺度適用中國國家揉準(CNS ) Α4规格(210X297公釐)L In the next cycle to the kth-Jth cycle, the multiplexers 724-0, 724_1 and 724_2 select the same sampling value, and the adders 726_〇 and 726-1 output the sum of the same sampling value. The selectors 735-0, 735-1 and 735-2 output the tap coefficient values corresponding to that cycle. That is, ----.----- ^ -installed ----- ^-ordered (please read the precautions on the back before filling in this page) This paper size is in accordance with Chinese national standards (CNS &gt; A4 specification (210X297 mm) 83.3.10,000 3807 (A7 B7) 5. Description of invention (4 left) Printed by the tec industrial and consumer cooperative of the Central Bureau of Economic Affairs of the Ministry of Economic Affairs in the second cycle, selector 735-0 outputs hi, selector 735-1 outputs h2 + h5, and selector 735_2 outputs 2h. Multipliers 740-0, 740-1, and 740-2 will correspond to the sampling value from the multiplexer 724-2 or from the adder 724_〇 and Sampling and multiplying 724 to generate individual row filter terms yx〇, yxi and yx2 0 L Starting from cycle-+ 1, the control signal Ct2 is changed to -2 ″ is 〇. This makes each multiplexer 724 -u (except the multiplexer 724-U) to select the sampling value stored in the register 722-nul as the input value of the adder 726-U. That is, the multiplexer 724_〇 has selected to store in 722- 3 The sampling value in the register and the multiplexer 724-1 select the sampling value stored in the temporary device 722_2. The multiplexer 724_2 selects the zero value. Therefore, the adder 726_0 will be stored in the register 722_0 and The sampling and output in 722-3, for example, χ (0) + χ (-3). The adder 726-1 will output the sum of the sampling values stored in the registers 722-1 and 722-2, namely χ (-ι) + χ (-2) 〇L In the (-J +1) th cycle, the selector 735-u outputs the coefficient value of the (-+ 1) th tap point stored in L. For example , In the second cycle of -2-, the selector 735-〇 output coefficient hi + h2, the selector 735-1 output coefficient h5 + h6 and the selector 735_2 output zero. Multiplication (please read the notes on the back before filling in this Page) This paper scale is applicable to the Chinese National Standard (CNS) A4 specification (210X297 mm) 83. 3. 10,000 V. Description of the invention (and) A7 B7 Printer 740-0 of the Zhengong Consumer Cooperative Society of the Central Bureau of Economic Development of the Ministry of Economic Affairs Then multiply the coefficient hi + h2 by the sum X (0) + X (_3). The multiplier 740-1 multiplies the coefficient h5 + h6 by the sum X (-l) + X (_2), and the multiplier will be The number zero is multiplied by zero. These products are treated as row filter terms yx〇, yh, and yx2, respectively, and output to the binary addition tree and accumulated to determine the value of yj (t). For each cycle, multiplexers 724_0, 724-1 and 724-2 are selected Zhi the same sampling and 726_1 and the adder 726-0 output and at the same sampling Zhi. The selectors 735-0, 735_1, and 735_2 output the tap coefficients relative to that loop. That is, in the fourth cycle, the selector 735-0 outputs hD + h3, and the selector 735-1 outputs h4 + h7, and the selector 735-2 outputs zero. The multipliers 740-0 and 740-1 are multiplied by the sum of the corresponding sampling values output from the adder 724-0 or 724-1 to generate individual modified row filter terms yxQ, yxi (multiplication The multiplier 74〇_2 multiplies the coefficient zero by the zero value output from the multiplexer 724_2 to generate the combined bank filter term yX2 = 0.) These combined bank filter terms yxQ, yXl, and yX2 are added in binary Tree 79〇 accumulated to determine the value of yj (t). Finally, after accumulating all J numbers of the output value yj (t) of the merge filter, the accumulator 750 outputs g (t). The above process is then repeated, that is, the next sampling value t-1 x (| ^ — J _i + l) (called, x (l)) is shifted to the previous place (please read the back side first) Note $ item and then fill out this page). 装 · Order _ This paper size is suitable for China National Standard (CNS) Α4 specification (210X297 mm) 83. 3.10,000 A7 B7 Printed by Beigong Consumer Cooperative Cooperative of Central Central Bureau of Economic Affairs System 5. Description of Invention (47) The shift register 722 of the management stage 710, and the signal ct2 is set to 1, and so on. At the beginning, the shift register 722 may be mostly empty. Therefore, the initial determination of * g (t) may not be very useful. Conversely, the shift register 722 can be pre-loaded with approximate sampling data so that all decision values of g (t) are available. The frequency multiplier 700 requires very low processing requirements compared to the FIR filter 400 (fifth figure). The binary adder tree 790 of the frequency multiplier 700 requires n adders to previously process the input sample value and less than (approximately 2 n ground)-adders used in the binary adder tree. The multiplication stage 730 requires 2 η. If-+ 1 multiplier is required, or an amount equal to the Booth multiplier is used by 2 n b, (-+ 1)-adders are required. Therefore, the 2 2 frequency multiplier according to the present invention can be executed with approximately ((ϋ + 1) .- ^-+ η) 2 adders. Conversely, the binary addition tree 43〇 of the finite impulse response (FIR) b filter 400 requires approximately (n + 1) -2 adders. Therefore, the traditional finite impulse response filter 40 (/ requires the following more adders: This paper size is applicable to the Chinese National Standard (CNS) A4 specification (210x297 mm) 83. 3.10,000 ^ — ^ 1, -... ..I n 1-— ^ ϋ In I (Please read the precautions on the back before filling in this page) Order A7 B7 V. Description of invention (fight?) Bn bnb (n + 1) ·--((-+ 1 ).-+ n) =-(--2) 2 2 2 2 2 Generally speaking, each video sampling value is more than 4 bits, that is, b &gt; 4 ° Therefore, use an upconverter 700 can achieve a saving effect. In short, the present invention discloses a multiplier with a combined filter bank by combining cross-symmetric filter banks. With this combination of filter banks, the processing process can be reduced Requirements (that is, the number of circuits, the space occupied by the circuits, and the processing time) without hindering the accuracy or efficiency of the frequency multiplier. However, the illustrations disclosed above in the present invention are only for the present invention. Embodiments, for those who are familiar with the skills of the case, according to the spirit of the case, the equivalent changes or modifications of the embodiment, It should be covered in the following patent applications in this case: ^^ 1 '— ^ 1 ^^ 1- ^^ 1 — ^ 1 · _ϋ— ^ (please read the precautions on the back before filling out this page) Order_Ministry of Economic Affairs Printed by the Beigong Consumer Cooperative of the Central Bureau of Standards 83. 3.10,000 This paper size is suitable for China National Standard (CNS) Α4 specification (210X297 mm)

Claims (1)

申請專利範圍 A8 B8 C8 D8 爷 —種數位濾波器,係依據L個濾波器排及线每一 濾波排的n.個分接點係數Pj? (i),其中;且 來產生一個輸入取樣^ x( t-l •i)的 L —種 L 相{ϋ插核器(L phase interpolation),該數 位濾波器包含有·: 一個第一個電路,用來產生,對每一個時問t, 一個合併排液波器的排涵!波器項(bankfilter terms), 其輸出値在前一個時問t-i及該時問t之間改變,线 產生排濾波器項包括至少兩個交叉對稱排濾波器項的 —種結合: 在處於同相位的情況下: t-k (請先閱讀背面之注意事項再填寫本頁) Pk (i)x( -i)+PL - x + ! (i) ' x( t-k L -i)及 L t-(L-K+l) _(n-l-i)) L 經濟部中央標準局員工消費合作社印製 在非同相位的情況下: t-k Pk (i) · x( i)+?L - k + i (i-l) · x( t-(L-K+l) L -1 L 本紙張尺度逋用中國國家標準(CNS ) Μ規格(210X297公釐) 申請專利範圍 t-k Pk (n-i) · χ( L A8 B8 C8 D8 -(n-i))+pL - κ + i (n-l'-i) . X( L 其中且 饮馼位濾波器所依據之:玟 L濾波器排的 任一非負整數w,t=w · L+K ;及 第二個電路,用來加總,對每一個時間t,在筏 i第一個電路中產生的it排濾波器項,以產生在酋 —時問t-1及該時問t之問變動的技合併排濾波 器輸出値。 2.如申請專利範圍第1項所述之馼位濾波器,其更包 含下列步驟: 一個第三個電路,用來加總,對每一個時間t, η合併排濾波器的輸出値,以產生該增頻値。 3 .如申請專利範圍第1項所述之數位濾波器,對每一 時間t, j第一個電路産生了前述之改變的合併 η (請先閲讀背面之注意事項再填寫本頁) 訂 -1 濾波器排的至多 個排濾波器項。 2 經濟部中央標準局員工消費合作社印裂 4·如申請專利範圆第2項所述之數位濾波器,對每 L 時間t 技第三個電'路將 2 合併排波波器輸 本纸張尺度適用中國國家標_( CNS ) A4規格(210X297公釐) 申請專利範圍 A8 B8 C8 D8 t- 出値相加,以產生 該輸入取樣値x( -i)的 L 一個增頻値g(t)。 5 ·如申請專利範圍第1項所述之數位濾波器,對一持 L 定時間t,使得-+1&lt;K&lt;L . 改變的合併濾 波器排輸出値如下: U-1 Σ 1=0 t-k t-k + 1 ⑴)· [χ( -i)+x( L L·」 匕Li」 -(n-i-l))] r- L (請先閱讀背面之注意事項再填寫本頁) 訂 其中η及L是偶數且 6 ·如申請專利範圍第1項所述之數位濾波器,對一持 L -ΙΛ 經濟部中央標準局員工消費合作社印製 定時問t,亦即I5_k三- 該改變的合併瀘波器排 2· 本紙張尺度適用中國國家標準(CNS ) A4規格(210X297公釐) 申請專利範圍 輸出値如下: yj (t)=pk (〇) . [X( U=2 L A8 B8 C8 D8 t-k -n)+x( L Σ (Ρχ (1)+Pl - κ + x (i-l)) · [X( t-k )] t-k t-k u L (請先閲讀背面之注意事項再填寫本頁) +2 · Pk (U-l) · x( -U^l) L 其中n及L爲偶·數,且 L 訂 • 一個具有合併瀘波器排的增頻器,其係藉由組合戌 增頻器的父叉對稱排濾波器項所形成,前述之增頻器包 含: 一個第一個電路,用來接收一系列自 t-i 經濟部中央標準局員工消費合作社印製 Χ( 幻至x( t-i 乙 -i-n)的取樣値,其 本紙張尺度適用中國國家標隼(CNS ) Α4規格(210X297公釐) 4 β07 β θ ABCD 申請專利範圍 中L·是 試增頻器的相位數,η是合併前的存 增頻器的每一濾波器排中的分接點係數量,且〇Si&lt; (η-l),及用來輸出一系列申結合形成交叉對稱排濾波 器項的取樣値; 一個第二個電路* '雨來輸出一系列分接點係數値, 其包括有前述之交叉對稱排濾波器項而需要分接點係數 的一個組舍;及 —個第二個電路,以接收茂..系列取樣値及夺 系列的分接點係數値的第三個電路,用來產生一個用 該系列取樣値的每一個取樣値與4. 系列分接點 係數値的相對應分接點係數値的乘稹,及經由誠 .乘 稹,產生一個增頻値。 8. 如申請專利範圍第7項所述之增頻器,择 第三個 電路,藉^ 一個自技 乘積其輸出値在前一個時間 t-l及一個現在(current)時間t間變動的合併排 .濾波器的一個輸出値,及藉由將該被決定的輸出値相 加至最少有一個其他合併排濾波器輸出値,其中最少 有一個在前一個時問t-Ι及現在時問t之問不會改 變輸出値,來決定增頓値。 9. 如申諳專利範圍第8項所述之增頻器,其中,絝 第三個電路包含有: 經濟部中央標準局員工消費合作社印製 (請先閱讀背面之注意事項再填寫本頁) —個累計器,闬來將該乘稹相加,以產生^ .已 改變的合併濾波器排輸出値,且用來將技 已改變 合併濾波器排輸出値加總至該至少有一個未改變的濾_ 波器排輸出値;及, 本纸張尺度適用中國國家標準(CNS ) Α4規格(210X297公釐) A8 B8 •C8 D8 中請.專利範圍 一個頻存器,用來儲存由前述之累計器產生的前述 之改蠻的合併濾波器排之輸出値,此輸出値在時間t 及下一個時間t+1之間不會改變。 ίο.如申請專利範圍第7項所述之增頻器,其中,前述之 第一個電路包含有: 一個隨機存取記憶體,用儲存前述之自 t-SL t-ί. x( -i)至 x( ~ L 」 —L 」 i-η)的取樣値;及 個加法器,用來將前述之隨機存取記憶體輸出的連續 取樣値相加在一起以產生前述之取樣値的結合。 11. 如申請專利範圍第10項所述之增頻器,其中,前述之 第一個電路包含有: 複數個計數器,用來爲前述之隨機存取記憶體定 .位址,及爲使該隨機存取記憶體能依照交叉對稱排濾 波器項連繪輸出取樣對(pairs of samples)。 12. 如申請專利範固第7項所述之增頻器,其中,前述之 第二電路包含有: 經濟部中央標準局員工消費合作社印製 (請先閲讀背面之注意事項再填寫本頁) —個唯讀記憶體,用來儲存前述之分接點係數及 該分接點係數組合;及 一個數器*在毎一個時問t,用來爲前述之唯 讀記憶體定位址,以便該唯詔記憶體輸出一個在前一 個時問t-l及該時問t之問改變値的合併濾波器排 的前述系列分接點係數値。 本紙張1適用中國國家標準(CNS) A4規格(210x297公釐) A8 B8 C8 D8 申請專利範園 13 .如申請專利範圍的第7項所述之增頻器,在每—個 時間t,前述之第一個電路執行至多個相 2 力η,及其中,前述之m三個電路執行至多 2 —Π _ -1個相加及 -+1個相乘,以産生每一個已 2 改變的合併濾波器排的輸出値。 14 ·如申請專利範圍第8項所述之增頻器,對每—時間 * |— L - t,前述之第三個電路執行至多——-1個合併渡 2 〜 ‘波器排輸出値的相加,以產生前述之增頻値。 15 .—個具有合併濾波器排的增頻器,其係藉由組合該增 頻器的交叉對稱排濾波器項所形成,此增頻器包含有 (請先閲讀背面之注意事項再填寫本頁) ·—ί 經濟部中夬標隼局員工消費合作社印製 本紙張尺度適用中國國家標準(CNS &gt; Μ規格(210&gt;&lt;297公釐) 申請專利範圍 A8 B8 C8 D8 —個丨吴組核心 (module core),用來接收—系列 t-1 t-l 自 | -| -i)至 x( L L •i_n)輸入取樣値 經濟部中央標準局員工消費合作社印製 ,其中L是前述增頻器的相位馼,n是該增頻器合 併前在每一個濾波器排中的分接點係數量,且 0&lt;i&lt;n -1,及一系列相對應的分接點係數値,這些係馼値包 括有一個交叉對稱排濾波器項組合的一個分接點係數 組合'前述之模組核心也用來平行地輸出,前述之合 併丨虔波器排的複數個排濾波器項,而該合併濾波器排 係藉由選擇性地將每一個該取樣値與一個相對應的分 接點係數値相乘所形成;及 —個相加樹電路,用來產生自前述的模組核心輸 出的前述之排濾波器項所輸出的一個合併排瀘波器的 .輸出値。 16 .如申請專利範圍第IS項所述之增頻器,更包含: 一個電路,用來儲存由前述之加法器電路所產生 的前述之合併排濾波器輸出値,及將此合併排濾波器 輸出値相加在一起,以產生一個增頻器取樣値。 17 .如申請專利範圍第15項所述之增頻器,其中,前述之 模組核心包含一個先前處理級,用來產生來自前述之 輸入取樣的複數個取樣値,其包括至少一個相對應於 前述分點係數的該取樣値組合。 (請先閱讀背面之注意事項再填寫本頁) 訂 本紙張尺度適用中國國家標準(CNS ) A4規格(210X297公釐) 六、申請專利範圍 經濟部中央標準局員工消費合作社印製 18 .如申請專利範圍第17項所述之增頻器,其中,前述之 先前處理級包含有: t -1 一個位移暫存器,用來儲存前述之自x( - -1) —L — 至x( - -i-n)輸入取樣値; —L - 複數個多工器,其包括至少一個多工器,用來將接收 到的儲存於前述位移暫存器中的兩個前述取樣値當作 是輸入値; 複數個加法器電路,其包括至少一個加法器電路, 用來將接收到的儲存於前述位移暫存器中的一個取樣 値及自前述之一個多工器所输出的一個取樣値,當作 是輸入値。_ 19.如申請專利範圍第17項所述之增頻器,其中,前述之 模組核心更包含: 複數個選擇器電路,用來儲存前述之分接點係數 値,及於特定循環時,平行地輸出儲存於其中之前述 分接點係數;及, 複馼個平行執行的乘法器,每一個乘法器依序地 將自前述之前處理級輸出的其中一取樣値中,與自前 述之選擇器電路輸出的一個相對應的前述之分接點係 數値相乘。 (請先閱讀背面之注意事項再填寫本頁) 訂 本紙張尺度適用中國國家標準(CNS ) A4規格(210X297公釐) 六、申請專利範圍 A8 B8 C8 D8 2〇·如申請專利範圍第15項所述之增頻器,其中,前述模 η 組核心包含用來將該取樣値相加在一起的—個加 2 η 法器電路,及用來産生合併濾波器排項的-+ 1 2 個乘法器電路,及其中前述之相加樹電路包含近似 η -個加法器電路。 2 21·如申請專利範圍第20項所述之增頻器,其中,乘法器. 電路是Booth的運算法則乘法器電路,其每一個電路 b 包含-個加法器電路,且其中該增頻器電路包含 2 n b 近似((-+1) · - +n)個加法器電路。 2 2 經濟部中央標準局員工消費合作社印裝 (請先閲讀背面之注意事項再填寫本頁)Patent application scope A8 B8 C8 D8-a digital filter, based on L filter rows and n. Tap coefficients Pj? (I) for each filter row of the line, where; and to generate an input sample ^ L of x (tl • i)-L phase {ϋ interpolation (L phase interpolation), the digital filter contains ·: a first circuit used to generate, for each time t, a merge The culvert of the draining wave filter! The bank filter terms (bankfilter terms), whose output value changes between the previous time ti and the time t, the line generating row filter term includes at least two crossed symmetric row filter terms -A combination of: in the same phase: tk (please read the notes on the back before filling this page) Pk (i) x (-i) + PL-x +! (I) 'x (tk L -i) and L t- (L-K + l) _ (nli)) L Printed by the employee consumer cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs in the case of different phases: tk Pk (i) · x (i) +? L-k + i (il) · x (t- (L-K + l) L -1 L This paper scale adopts the Chinese National Standard (CNS) Μ specification (210X297mm). The scope of patent application tk Pk (n -i) · χ (L A8 B8 C8 D8-(ni)) + pL-κ + i (n-l'-i). X (L and the basis of the position filter: Min L filter row Any non-negative integer w, t = w · L + K; and the second circuit is used to sum up, for each time t, the row of filter items generated in the first circuit of raft i to produce The output value of the technical alignment filter that changes at the time t-1 and the time t. 2. As described in item 1 of the patent application scope, it includes the following steps: a Three circuits are used to sum up, and at each time t, η combine the output values of the filter to produce the frequency-increasing value. 3. The digital filter as described in item 1 of the patent application scope, for each At time t, j the first circuit produced the merger of the aforementioned changes η (please read the precautions on the back before filling in this page). Order -1 filter row to multiple row filter items. 2 Central Bureau of Economics employees Printed by the consumer cooperative 4. The digital filter as described in item 2 of the patent application, for the third circuit every L time t, the 2 combined wave filter output This paper scale is applicable to the Chinese national standard _ (CNS) A4 specification (210X297 mm). Patent scope A8 B8 C8 D8 t- The output value is added to generate the input sampling value x (-i) L An increased frequency value g (t). 5 · For the digital filter described in item 1 of the patent application scope, hold L for a fixed time t such that-+ 1 &lt; K &lt; L. The output value of the combined filter row changed is as follows: U-1 Σ 1 = 0 tk tk + 1 ⑴) · [χ (-i) + x (LL · "Dagger Li"-(nil))] r- L (please read the precautions on the back and then fill out this page), where η and L are Even number 6 · The digital filter as described in item 1 of the patent application scope is printed on the time when a consumer ’s consumer cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs, namely, I5_k III-the changed merger filter Row 2. This paper scale is applicable to the Chinese National Standard (CNS) A4 specification (210X297 mm). The output value of the patent application range is as follows: yj (t) = pk (〇). [X (U = 2 L A8 B8 C8 D8 tk- n) + x (L Σ (Ρχ (1) + Pl-κ + x (il)) · [X (tk)] tk tk u L (please read the precautions on the back before filling this page) +2 · Pk (Ul) x (-U ^ l) L where n and L are even numbers, and L is set • a frequency multiplier with a combined wave bank, which is symmetrical by combining the parent fork of the frequency multiplier Row filter As a result, the aforementioned frequency upconverter includes: a first circuit for receiving a series of sampling values printed by Χ (幻 至 x (ti 乙 -in) printed by the employee consumer cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs. The paper scale is applicable to China National Standard Falcon (CNS) Α4 specification (210X297 mm) 4 β07 β θ ABCD In the patent application scope, L · is the phase number of the trial multiplier, η is each filter of the pre-merger multiplier The number of tap coefficients in the device bank, and 〇Si &lt; (η-l), and the sampling value used to output a series of applications combined to form a cross-symmetric filter bank; a second circuit * 'Rain to output a Series tap coefficients, which include a group that requires the tap coefficients of the aforementioned cross-symmetrical row filter term; and-a second circuit to receive the serial sampling value and the series of points The third circuit of the contact coefficient value is used to generate a multiplication of each sampling value of the series sampling value and the corresponding tap coefficient value of the 4. series tap coefficient value, and by Cheng. Multiplication Zhen, produces an increased frequency. 8. As the frequency increaser described in item 7 of the patent application, select the third circuit, and use ^ a self-technical product whose output value changes between the previous time tl and a current (current) time t. An output value of, and by adding the determined output value to the output value of at least one other merger filter, at least one of which asks t-Ι in the previous one and asks t in the present not Change the output value to determine the increase value. 9. The frequency increaser as described in Item 8 of the patent scope, in which the third circuit includes: Printed by the Consumer Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs (please read the precautions on the back before filling this page) An accumulator that adds the multiplier to produce ^. The changed merge filter row output value, and is used to add the changed merge filter row output value to at least one unchanged The output value of the filter_wave filter; and, the paper scale is applicable to the Chinese National Standard (CNS) Α4 specification (210X297mm) A8 B8 • C8 D8. Please patent a frequency memory, used to store the aforementioned The output value of the combined filter bank generated by the accumulator described above does not change between time t and the next time t + 1. ίο. The frequency increaser as described in item 7 of the patent application scope, wherein the aforementioned first circuit includes: a random access memory for storing the aforementioned self-t-SL t-ί. x (-i ) To x (~ L ”—L” i-η); and an adder used to add the aforementioned continuous sampling values of the random access memory output to produce a combination of the aforementioned sampling values . 11. The frequency increaser as described in item 10 of the patent application scope, wherein the aforementioned first circuit includes: a plurality of counters, which are used to address the aforementioned random access memory, and to enable the The random access memory can continuously output pairs of samples according to the cross-symmetry filter term. 12. The frequency increaser as described in item 7 of the patent application Fan Gu, in which the aforementioned second circuit includes: Printed by the Consumer Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs (please read the precautions on the back before filling this page) -A read-only memory, used to store the aforementioned tap coefficients and the tap-point coefficient combinations; and a counter * at each time, t is used to locate the aforementioned read-only memory so that the Only the memory outputs a series of tap point coefficient values of the merge filter row that changed the value at the previous time t and the time t. This paper 1 is applicable to the Chinese National Standard (CNS) A4 specification (210x297 mm) A8 B8 C8 D8 Patent Application Fan Garden 13. As described in item 7 of the scope of the patent application frequency increaser, at each time t, the aforementioned The first circuit executes up to multiple phase 2 forces η, and among them, the aforementioned m three circuits execute up to 2 —Π -1 addition and-+ 1 multiplication to produce each merger that has 2 changes The output value of the filter bank. 14 · The frequency increaser as described in item 8 of the patent application scope, for every-time * |-L-t, the aforementioned third circuit is executed at most--1 merger 2 ~ 'waveform row output value To add the aforementioned frequency increase value. 15. A multiplier with a combined filter bank is formed by combining the cross-symmetrical bank filter items of the multiplier. This multiplier contains (please read the precautions on the back before filling in this Page) · —The paper printed by the Ministry of Economic Affairs Zhongshen Standard Falcon Bureau Employee Consumer Cooperative applies the Chinese national standard (CNS &gt; M specifications (210 &gt; &lt; 297mm). Patent scope A8 B8 C8 D8-a 丨 Wu Module core, used to receive-series t-1 tl from |-| -i) to x (LL • i_n) input sampling value Printed by the Employee Consumer Cooperative of the Central Standards Bureau of the Ministry of Economy, where L is the aforementioned frequency increase The phase of the filter, n is the amount of tap coefficients in each filter bank before the frequency multiplier is combined, and 0 &lt; i &lt; n -1, and a series of corresponding tap coefficients, these systems馼 値 includes a combination of tap coefficients with a cross-symmetric row filter term combination. The aforementioned module core is also used to output in parallel. The combined filter row is The sampling value is formed by multiplying a corresponding tap coefficient value; and an addition tree circuit is used to generate a combined row output from the aforementioned row filter term output from the aforementioned module core The output value of the wave filter. 16. The frequency multiplier as described in item IS of the patent application scope further includes: a circuit for storing the output value of the aforementioned combined bank filter generated by the aforementioned adder circuit, and the combined bank filter The output values are added together to produce a multiplier sampling value. 17. The frequency multiplier as described in item 15 of the patent application scope, wherein the aforementioned module core includes a previous processing stage for generating a plurality of sampling values from the aforementioned input samples, including at least one corresponding to This sampling value combination of the aforementioned dividing point coefficients. (Please read the precautions on the back before filling in this page) The size of the revised paper is in accordance with the Chinese National Standard (CNS) A4 (210X297 mm). 6. Scope of Patent Application Printed by the Employee Consumer Cooperative of the Central Standards Bureau of the Ministry of Economy. The frequency increaser described in item 17 of the patent scope, wherein the aforementioned previous processing stage includes: t -1 a shift register for storing the aforementioned from x (--1) —L — to x (- -in) input sampling value; -L-a plurality of multiplexers, including at least one multiplexer, used to receive the two aforementioned sampling values stored in the displacement register as input values; A plurality of adder circuits, including at least one adder circuit, for receiving a sample value stored in the shift register and a sample value output from the multiplexer, as Enter the value. _ 19. The frequency increaser as described in item 17 of the patent application scope, wherein the aforementioned module core further includes: a plurality of selector circuits for storing the aforementioned tap point coefficient values, and at specific cycles, The aforementioned tap coefficients stored therein are output in parallel; and, a plurality of multipliers executed in parallel, each multiplier sequentially selects one of the sampled values output from the previous processing stage, and selects from the aforementioned The output of the converter circuit is multiplied by a corresponding tap point coefficient value. (Please read the precautions on the back before filling in this page) The size of the revised paper applies to the Chinese National Standard (CNS) A4 (210X297mm) 6. Scope of patent application A8 B8 C8 D8 2 The frequency multiplier described above, wherein the aforementioned modular η group cores include an adder 2 η adder circuit for adding the sampling values together, and − + 1 2 for generating a combined filter row The multiplier circuit and the aforementioned addition tree circuit include approximately n-adder circuits. 2 21. The frequency multiplier as described in item 20 of the patent application scope, wherein the multiplier. The circuit is Booth ’s algorithm multiplier circuit, each circuit b of which includes an adder circuit, and wherein the frequency multiplier The circuit contains 2 nb approximate ((-+ 1) ·-+ n) adder circuits. 2 2 Printed by the Employees Consumer Cooperative of the Central Bureau of Standards of the Ministry of Economic Affairs (please read the precautions on the back before filling this page) 本纸張尺度適用中國國家標準(CNS ) A4規格(210X297公釐)This paper scale is applicable to China National Standard (CNS) A4 specification (210X297mm)
TW83108068A 1994-08-30 1994-08-30 Signal processing method and device of interpolation digital filter TW320794B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
TW83108068A TW320794B (en) 1994-08-30 1994-08-30 Signal processing method and device of interpolation digital filter

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
TW83108068A TW320794B (en) 1994-08-30 1994-08-30 Signal processing method and device of interpolation digital filter

Publications (1)

Publication Number Publication Date
TW320794B true TW320794B (en) 1997-11-21

Family

ID=51567067

Family Applications (1)

Application Number Title Priority Date Filing Date
TW83108068A TW320794B (en) 1994-08-30 1994-08-30 Signal processing method and device of interpolation digital filter

Country Status (1)

Country Link
TW (1) TW320794B (en)

Similar Documents

Publication Publication Date Title
US5287299A (en) Method and apparatus for implementing a digital filter employing coefficients expressed as sums of 2 to an integer power
JP3000325B2 (en) Finite impulse response digital filter
WO1996023353A1 (en) Finite impulse response filter
JP2916056B2 (en) Matrix multiplication circuit
KR0181165B1 (en) Impulse response filter and filtering method
WO1999038088A1 (en) Method and apparatus for arithmetic operation
Wang et al. Moduli selection in RNS for efficient VLSI implementation
JP2003516692A (en) Reconfigurable finite response filter using canonical sign digit number coefficient representation
JP2008124593A (en) Decimation filter
KR100295257B1 (en) Digital filter
CN113556101B (en) IIR filter and data processing method thereof
JP2002522939A (en) Digital filtering without multiplier
JPH0377418A (en) Digital filter
US5016011A (en) Increased performance of digital integrated circuits by processing with multiple-bit-width digits
TW320794B (en) Signal processing method and device of interpolation digital filter
Yazhini et al. Fir filter implementation using modified distributed arithmetic architecture
US7728743B2 (en) Device and method for polyphase resampling
CN101840322B (en) The arithmetic system of the method that filter arithmetic element is multiplexing and wave filter
RU2015550C1 (en) Arithmetic unit for performing discrete fouler transform
US10305451B1 (en) Multiplier-based programmable filters
Bistritz et al. Inversion and factorization of non-Hermitian quasi-Toeplitz matrices
KR100235537B1 (en) Variable tap of digital filter and multiplier circuit thereof
JP3197648B2 (en) Digital filter
JPH05175785A (en) Digital filter for decimation
JP2864597B2 (en) Digital arithmetic circuit

Legal Events

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