Cyclic Codes

Yunghsiang S. Han

Graduate Institute of Communication Engineering, National Taipei University Taiwan E-mail: [email protected]

Y. S. Han

Cyclic codes

Description of Cyclic Codes

If the components of an n-tuple v = (v0 , v1 , . . . , vn1 ) are cyclically shifted i places to the right, the resultant n-tuple would be v (i) = (vni , vni+1 , . . . , vn1 , v0 , v1 , . . . , vni1 ). Cyclically shifting v i places to the right is equivalent to cyclically shifting v n i places to the left. An (n, k ) linear code C is called a cyclic code if every cyclic shift of a code vector in C is also a code vector in C . Code polynomial v (x) of the code vector v is dened as v (x) = v0 + v1 x + + vn1 xn1 . v (i) (x) = xi v (x) mod xn + 1.
Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

Proof: Multiplying v (x) by xi , we obtain xi v (x) = v0 xi + v1 xi+1 + + vni1 xn1 + + vn1 xn+i1 . Then we manipulate the equation into the following form: xi v (x) = vni + vni+1 x + + vn1 xi1 + v0 xi + +vni1 xn1 + vni (xn + 1) + vni1 x(xn + 1) + + vn1 xi1 (xn + 1) = q (x)(xn + 1) + v (i) (x), where q (x) = vni + vni+1 x + + vn1 xi1 . The nonzero code polynomial of minimum degree in a cyclic code C is unique. Let g (x) = g0 + g1 x + + gr1 xr1 + xr be the nonzero code polynomial of minimum degree in an (n, k ) cyclic code C . Then
Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

the constant term g0 must be equal to 1. Proof: Suppose that g0 = 0. Then g (x) = = g1 x + g2 x2 + + gr1 xr1 + xr x(g1 + g2 x + + gr1 xr2 + xr1 ).

If we shift g (x) cyclically n 1 places to the right (or one place to the left), we obtain a nonzero code polynomial, g1 + g2 x + + gr1 xr2 + xr1 , which has a degree less than r . Contradiction.

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

A (7, 4) Cyclic Code Gnerated by g (x) = 1 + x + x3

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

Consider the polynomial xg (x), x2 g (x), . . . , xnr1 g (x). Clearly, they are cyclic shifts of g (x) and hence code polynomials in C . Since C is linear, a linear combination of g (x), xg (x), . . . , xnr1 g (x), v (x) = u0 g (x) + u1 xg (x) + + unr1 xnr1 g (x) = (u0 + u1 x + + unr1 xnr1 )g (x),

is also a code polynomial where ui {0, 1}. Let g (x) = 1 + g1 x + + gr1 xr1 + xr . be the nonzero code polynomial of minimum degree in an (n, k ) cyclic code C . A binary polynomial of degree n 1 or less is a code polynomial if and only if it is a multiple of g (x). Proof: Let v (x) be a binary polynomial of degree n 1 or less.

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

Suppose that v (x) is a multiple of g (x). Then v (x) = (a0 + a1 x + + anr1 xnr1 )g (x) = a0 g (x) + a1 xg (x) + + anr1 xnr1 g (x).

Since v (x) ia a linear combination of the code polynomials, g (x), xg (x), . . . , xnr1 g (x), it is a code polynomial in C . Now let v (x) be a code polynomial in C . Dividing v (x) by g (x), we obtain v (x) = a(x)g (x) + b(x), where the degree of b(x) is less than the degree of g (x). Since v (x) and a(x)g (x) are code polynomials, b(x) is also a code polynomial. Suppose b(x) = 0. Then b(x) is a code polynomial with less degree than that of g (x). Contradiction. The number of binary polynomials of degree n 1 or less that are multiples of g (x) is 2nr .
Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

There are total of 2k code polynomials in C , 2nr = 2k , i.e., r = n k. The polynomial g (x) is called the generator polynomial of the code. The degree of g (x) is equal toi the number of parity-check digits of the code. The generator polynomial g (x) of an (n, k ) cyclic code is a factor of xn + 1. Proof: We have xk g (x) = (xn + 1) + g (k) (x). Since g (k) (x) is the code polynomial obtained by shifting g (x) to
Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

the right cyclically k times, g (k) (x) is a multiple of g (x). Hence, xn + 1 = {xk + a(x)}g (x). If g (x) is a polynomial of degree n k and is a factor of xn + 1, then g (x) generates an (n, k ) cyclic code. Proof: A linear combination of g (x), xg (x), . . . , xk1 g (x), v (x) = = a0 g (x) + a1 xg (x) + + ak1 xk1 g (x) (a0 + a1 x + + ak1 xk1 )g (x),

is a polynomial of degree n 1 or less and is a multiple of g (x). There are a total of 2k such polynomial and they form an (n, k ) linear code. Let v (x) = v0 + v1 x + + vn1 xn1 be a code polynomial in

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

this code. We have xv (x) = = = v0 x + v1 x2 + + vn1 xn vn1 (xn + 1) + (vn1 + v0 x + + vn2 xn1 ) vn1 (xn + 1) + v (1) (x).

Since both xv (x) and xn + 1 are divisible by g (x), v (1) must be divisible by g (x). Hence, v (1) (x) is a code polynomial and the code generated by g (x) is a cyclic code. Suppose that the message to be encoded is u = (u0 , u1 , . . . , uk1 ). Then xnk u(x) = u0 xnk + u1 xnk+1 + + uk1 xn1 . Dividing xnk u(x) by g (x), we have xnk u(x) = a(x)g (x) + b(x).
Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

10

Since the degree of g (x) is n k , the degree of b(x) must be n k 1 or less. Then b(x) + xnk u(x) = a(x)g (x) is a multiple of g (x) and therefore it is a code polynomial. b(x) + xnk u(x) = b0 + b1 x + + bnk1 xnk1 +u0 xnk + u1 xnk+1 + + uk1 xn1 then corresponds to the code vector (b0 , b1 , . . . , bnk1 , u0 , u1 , . . . , uk1 ).

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

11

A (7, 4) Cyclic Code Gnerated by g (x) = 1 + x + x3

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

12

Generator and Parity-Check Matrices

The generator matrix of an (n, k ) code C is as follows:
g0 0 0 G = 0 g1 g0 0 g2 g1 g0 g2 g1 g2 gnk 0 gnk 0 0 gnk 0 0 0 0 0 0 0 0 g0 g1 g2 gnk

In general, G is not in systematic form. However, it can be put into systematic form with row operation. Let xn + 1 = g (x)h(x),

where the polynomial h(x) has the degree k and is of the following form: h(x) = h0 + h1 x + + hk xk
Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

13

with h0 = hk = 1. A parity-check matrix of C may be obtained from h(x). Let v be a code vector in C and v (x) = a(x)g (x). Then v (x)h(x) = = = a(x)g (x)h(x) a(x)(xn + 1) a(x) + xn a(x).

Since the degree of a(x) is k 1 or less, the powers xk , xk+1 , . . . , xn1 do not appear in a(x) + xn a(x). Therefore,
k

hi vnij = 0 for 1 j n k.
i=0

We take the reciprocal of h(x), xk h(x1 ) = hk + hk1 x + hk2 x2 + + h0 xk ,

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

14

and can see that xk h(x1 ) is also a factor of xn + 1. xk h(x1 ) then generates an (n, n k ) cyclic code with the following (n k ) n matrix as a generator matrix:
H = hk 0 0 0 0 0 hk hk1 hk2 hk1 hk 0 hk2 hk1 hk hk2 hk1 h0 0 h0 0 0 h0 0 0 0 h0

Then H is a parity-check matrix of the cyclic code C . We call h(x) the parity polynomial of C . Let C be an (n, k ) cyclic code with generator polynomial g (x). The dual code of C is also cyclic and is generated by the polynomial xk h(x1 ), where h(x) = (xn + 1)/g (x). Let xnk1 = ai (x)g (x) + bi (x) for 0 i k 1,
Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

15

where bi (x) = bi0 + bi1 + + bi(nk1) . Since bi (x) + xnk+i are multiples of g (x), they are code polynomials. Then
G = b00 b10 b20 b01 b11 b21 . . . b(k1)0 b(k1)1 b(k1)2 b(k1)(nk1) b02 b12 b22 b0(nk1) b1(nk1) b2(nk1) 1 0 0 . . . 0 0 0 0 1 0 0 1 0 0 0 . . . 1 .

The corresponding parity-check matrix for C is

H = 1 0 0 0 1 0 . . . 0 0 0 1 0 0 1 0 0 0 b00 b01 b02 . . . b0(nk1) b1(nk1) b2(nk1) b10 b11 b12 b20 b21 b22 b(k1)0 b(k1)1 b(k1)2 . . . b(k1)(nk1) .

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

16

Encoding of Cyclic Codes

Encoding process: (1) Multiply u(x) by xnk ; (2) divide xnk u(x) by g (x); (3) form the code word b(x) + xnk u(x).
( ) ( ) g1 b0 b1 n-k stage g2 b2 ( ) Gate gn-k-1 bn-k-1 Code word Parity-check digits

Message Xn-ku(X)

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

17

Example
Consider the (7, 4) cyclic code generated by g (x) = 1 + x + x3 . Suppose that the message u = (1 0 1 1) is to be encoded. The contentents in the register are as follows:
Input 1 1 0 1 Register contents 0 0 0 (initial state) 1 1 0 (first shift) 1 0 1 (second shift) 1 0 0 (third shift) 1 0 0 (fourth shift)

After four shifts, the contents of the register are (1 0 0). Thus the complete code vector is (1 0 0 1 0 1 1).
Gate

(1 0 1 1)

Message Xn-ku(X)

Parity digits

Code word

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

18

Encoding by Parity Polynomial

Since hk = 1, we have
k 1

vnkj =
i=0

hi vnij for 1 j n k,

which is known as a dierence equation. vnk1 = h0 vn1 +h1 vn2 + +hk1 vnk = uk1 +h1 uk2 + +hk1 u0 vnk2 = uk2 + h1 uk3 + + hk1 vnk1 Encoding circuit:
Gate2 Gate1

+
hk-1

+
hk-2

+
h2

+
h1

u0

u1

Uk-2 k stage

Uk-1

Output to channel

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

19

Example
The parity polynomial of the (7, 4) cyclic code generated by g (x) = 1 + x + x3 is x7 + 1 2 4 = 1 + x + x + x . h(x) = 3 1+x+x The encoding circuit:
h0 = 1 Input Gate 1 Gate 2 p Output

Suppose that the message to be encoded is (1 0 1 1). Then v3 = 1, v4 = 0, v5 = 1, v6 = 1. The parity-check digits are v2 v1 = = v6 + v3 + v4 = 1 + 1 + 0 = 0 v5 + v4 + v3 = 1 + 0 + 1 = 0

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

20

v0

v4 + v3 + v2 = 0 + 1 + 0 = 1.

The code vector that corresponds to the message (1 0 1 1) is (1 0 0 1 0 1 1).

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

21

Syndrome Computation
Let r = (r0 , r1 , . . . , rn1 ) be the received vector. The syndrome is calculated as s = r H T , where H is the parity-check matrix. If syndrome is not identical to zero, r is not a code vector and the presence of errors has been detected. Dividing r (x) by the generator polynomial g (x), we obtain r(x) = a(x)g (x) + s(x). The n k coecients of s(x) form the syndrome s. We call s(x) Gate the syndrome.
r(x)/g(x) r(X)
Received vector

g1 s0 s1

g2

gn-k-1

sn-k-1

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

22

If C is a systematic code, then the syndrome is simply the vector sum of the received parity digits and the parity-check digits recomputed from the received information digits. Let s(x) be the syndrome of a received polynomial r (x). Then the remainder s(1) (x) resulting from dividing xs(x) by the generator polynomial g (x) is the syndrome of r (1) (x), which is a cyclic shift of r (x). Proof: We have xr (x) = rn1 (xn + 1) + r(1) (x). Then c(x)g (x) + (x) = rn1 g (x)h(x) + x[a(x)g (x) + s(x)], where (x) is the remainder resulting from dividing r(1) (x) by g (x). Then (x) is the syndrome of r (1) (x). Rearranging the
Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

23

above equation, we have xs(x) = [c(x) + rn1 h(x) + xa(x)]g (x) + (x). It is clearly that (x) is also the remainder resulting from dividing xs(x) by g (x). Therefore, (x) = s(1) (x). The remainder s(i) (x) resulting from dividing xi s(x) be the generator polynomial g (x) is the syndrome of r (i) (x), which is the ith cyclic shift of r (x).

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

24

Example
Consider the (7, 4) cyclic code generated by g (x) = 1 + x + x3 . Suppose that the received vector is r = (0 0 1 0 1 1 0). The syndrome of r is s = (1 0 1). As the received vector is shifted into the circuit, the contents in the register are as follows:
Gate input Gate
1 0 1

If the register is shifted once more with the input gate disabled, the new contents will be s(1) = (1 0 0), which is the syndrome of r(1) = (0 0 0 1 0 1 1).
Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

25

We may shift the received vector r(x) into the syndrome register from the right end. However, after the entire r (x) has been shifted into the register, the contents in the register do not form the sybdrome of r (x); rather, they form the syndrome s(nk) (x) of r(nk) (x).
Gate xn-kr(x)/g(x) s0 g1 s1 g2
Received vector

gn-k-1

sn-k-1 r(X)

Proof: We have xnk r(x) = a(x)g (x) + (x).

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

26

It is known that xnk r (x) = b(x)(xn + 1) + r(nk) (x). Hence, r (nk) (x) = [b(x)h(x) + a(x)]g (x) + (x). When r (nk) (x) is divided by g (x), (x) is also the remainder. Therefore, (x) is indeed the syndrome of r (nk) (x).

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

27

Error Detection
Let v (x) be the transmitted code word and e(x) = e0 + e1 x + + en1 xn1 be the error pattern. Then r (x) = v (x) + e(x) = b(x)g (x) + e(x). Following the denition of syndrome, we have e(x) = [a(x) + b(x)]g (x) + s(x). This shows that the syndrome is actually equal to the remainder resulting from dividing the error pattern by the generator polynomial. The decoder has to estimate e(x) based on the syndrome s(x). If e(x) is identical to a code vector, e(x) is an undetectable error pattern.
Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

28

The error-detection circuit is simply a syndrome circuit with an OR gate with the syndrome digits as inputs. For a cyclic code, an error pattern with errors conned to i high-order positions and i low-order positions is also regarded as a burst of length or less. such a burst is called end-around burst. An (n, k ) cyclic code is capable of detecting any error burst of length n k or less, including the end-around bursts. Proof: Suppose that the error pattern is a burst of length of n k or less. Then e(x) = xj B (x), where 0 j n 1 and B (x) is a polynomial of degree n k 1 or less. Since the degree of B (x) is less than that of g (x), B (x) is not divisible by g (x). Since g (x) is a factor of
Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

29

xn + 1 and x is not a factor of g (x), g (x) and xj must be relatively prime. Therefore, e(x) is not divisible by g (x). The last part of the above statement is left as an exercise. The fraction of undetectable bursts of length n k + 1 is 2(nk1) . Proof: Consider the bursts of length n k + 1 starting from the ith digit position and ending at the (i + n k )th digit position. There are 2nk1 such burst. Among these bursts, the only one that cannot be detected is e(x) = xi g (x). Therefore, the fraction of undetectable bursts of length n k + 1 starting from the ith digit position is 2(nk1) . For > n k + 1, the fraction of undetectable error bursts of length is 2(nk) . The proof is left as an exercise.
Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

30

Decoding of Cyclic Codes

Decoding of linear codes consists of three steps: (1) syndrome computation; (2) association of the syndrome to an error pattern; (3) error correction. The cyclic structure structure of a cyclic code allows us to decode a received vector r (x) in serial manner. The received digits are decoded one at a time and each digit is decoded with the same circuitry. The decoding circuit checks whether the syndrome (x) corresponds to a correctable error pattern e(x) with an error at the highest-order position xn1 (i.e., en1 = 1). If (x) does not correspond to an error pattern with en1 = 1, the received polynomial and the syndrome register are cyclically shifted once simultaneously. By doing this, we have r (1) (x) and
Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

31

s(1) (x). The second digit rn2 of r (x) becomes the rst digit of r (1) (x). The same decoding processes. If the syndrome s(x) of r (x) does correspond to an error pattern with an error at the location xn1 , the rst received digit rn1 is an erroneous digit and it must be corrected by taking the sum rn1 en1 . This correction results in a modied received polynomial, denoted by r1 (x) = r0 + r1 x + + rn2 xn2 + (rn1 en1 )xn1 . The eect of the error digit en1 on the syndrome can be achieved by adding the syndrome of e (x) = xn1 to s(x). The syndrome s1 of r1 (x) is the remainder resulting from dividing x[s(x) + xn1 ] by the generator polynomial g (x).
Graduate Institute of Communication Engineering, National Taipei University

(1)

(1)

Y. S. Han

Cyclic codes

32

Since the remainders resulting from dividing xs(x) and xn by g (x) are s(1) (x) and 1, respectively, we have s1 (x) = s(1)(x) + 1.
(1)

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

33

Meggitt Decoder I
Gate (r0, r1, , rn-1)
r(X) Received vector Corrected vector

Gate

Buffer registers Feedback connection Gate

ri

Gate Syndrome register s(x) --- r(x) s(1)(x) --- r(1)(x) ei Gate

Error pattern detection circuit Syndrome modification

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

34

Example
Consider the decoding of the (7, 4) cyclic code generated by g (x) = 1 + x + x3 . This code has minimum distance 3 and is capable of correcting any single error. The seven single-error patterns and their corresponding syndromes are as follows:
Error pattern e(X) e6(X) = X6 e5(X) = X5 e4(X) = X4 e3(X) = X3 e2(X) = X2 e1(X) = X1 e0(X) = X0 Syndrome s(X) s(X) = 1 + X2 s(X) = 1 + X + X2 s(X) = X + X2 s(X) = 1 + X s(X) = X2 s(X) = X s(X) = 1 Syndrome vector (s0, s1, s2) (1 0 1) (1 1 1) (0 1 1) (1 1 0) (0 0 1) (0 1 0) (1 0 0)

Suppose that the code vector v = (1 0 1 1 0 1 1) is transmitted and r = (1 0 1 1 0 1 1).

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

35

Buffer register r(X) Input Gate Gate 0 0 1 Multiplexer

r(X) Output

Gate

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

36

Syndrome register

Buffer register Pointer 1 0 1 1 0 1 1

Correction 0

Initial

0 0 1

0 1st shift 1 1 0 1 1 0 1 1 0 1 0 2st shift 0 1 1 1 1 1 0 1 1 0 0 3rd shift 1 1 1 0 1 1 1 0 1 1

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

37

1 4th shift 1 0 1 1 0 1 1 1 0 1 0

Error corrected 5th shift 0 0 0 0 1 0 1 1 1 0

0 6th shift 0 0 0 0 0 1 0 1 1 1 0

Corrected word 7th shift 0 0 0 1 0 0 1 0 1 1

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

38

Meggitt Decoder II
To decode a cyclic code, the received polynomial r (x) may be shifted into the syndrome register from the right end for computing the syndrome. When r (x) has been shifted into the syndrome register, the register contains s(nk) (x), which is the syndrome of r(nk) (x). If s(nk) (x) corresponds to an error pattern e(x) with en1 = 1, the highest-order digit rn1 of r (x) is erroneous and must be corrected. In r(nk) (x), the digit rn1 is at the location xnk1 . When rn1 is corrected, the error eect must be removed from s(nk) (x). The new syndrome s1 (x) is the sum of s(nk) (x) and the remainder (x) resulting from dividing xnk1 by g (x). Since
Graduate Institute of Communication Engineering, National Taipei University

(nk)

Y. S. Han

Cyclic codes

39

the degree of xnk1 is less than the degree of g (x), s1

(r0, r1, ,rn-1) r(X) Received vector =
(nk)

(x) = s(nk) (x) + xnk1 .

Gate ri Gate
Corrected vector

Gate

Buffer register r(X)

Feedback connection

Gate

Syndrome register

(Syndrome modification)

Error-pattern detection circuit

ei

Gate

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

40

Example
Again, we consider the decoding of the (7, 4) cyclic code generated by g (X ) = 1 + X + X 3 . Suppose that the received polynomial r (X ) is shifted into the syndrome register from the right end. The seven single-error patterns and their corresponding syndromes are as follows:
Error pattern e(X) e(X) = X6 e(X) = X5 e(X) = X4 e(X) = X3 e(X) = X2 e(X) = X1 e(X) = X0 Syndrome s(3)(X) s(3)(X) = X2 s(3)(X) = X s(3)(X) = 1 s(3)(X) = 1 + X2 s(3)(X) = 1 + X + X2 s(3)(X) = X + X2 s(3)(X) = 1 + X Syndrome vector (s0, s1, s2) (0 0 1) (0 1 0) (1 0 0) (1 0 1) (1 1 1) (0 1 1) (1 1 0)

We see that only when e(X ) = X 6 occurs, the syndrome is (0 0 1)

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

41

after the entire received polynomial r (X ) has been shifted into the syndrome register. If the single error occurs at the location X i with i = 6, the syndrome in the register will not be (0 0 1) after the entire received polynomial r (X ) has been shifted into the syndrome register. However, another 6i shifts, the syndrome register will contain (0 0 1). Based on this fact, we obtain another decoding circuit for the (7, 4) cyclic code generated by g (X ) = 1 + X + X 3 .

Graduate Institute of Communication Engineering, National Taipei University

Y. S. Han

Cyclic codes

42

r(X) Input

Buffer register Multiplexer

r(X) Output

Gate Gate

Gate

Graduate Institute of Communication Engineering, National Taipei University