5.1 INTRODUCTION In unit 2, we discussed various lower value of average length, thereby increasing the coding efficiency. The disadvantage with this type of coding is that they are “variable-length? codes, Due to this, a single error which occurs due to the noise present in the channel, affects more than one block-code- words. Another disadvantage of Variable-length codes is that the ‘output data rates measured ever short time-periods will fluctuate wid used, a single error will affect only that block which can be easily detected and corrected, To detect and comect errors, we go in for “error-control coding ing” techniques that rely on the systematic addition of “redundant” symbols. In this chapter, let us discuss i necessity of exror control coding, in detail, the exact meaning of error control coding, the and also the various ways of achieving it, 52 RATIONALE FOR CODING AND TYPES OF CODES The two key system parameters available in desi igning acost effective and reliable digital communication system, are “signal power” and “channel bandwidth” These two, along with PSD of noise “n” determine the bit signal energy to noise power ratio (E,/N). This ratio, intum, determines the bit error rate for various digital modulation schemes. Practical aspects place a limit on the value of (E,/N) [Refer section 4.6]. In practice, we find that its impossible ‘0 provide the acceptable data quality with whatever modulation schemes that we adopt. Hence, the only practical option available to improve the data quality is “error control coding”. Error control coding is nothing but calculated use of “redundancy”. The functional blocks thataccomplish error control coding are the “channel encoder” atthe transmitter and “channel ‘oder” at the receiver. For this reason error control coding is also termed as “channel ‘coding”’, Error control coding improves the data quality to a great orient ee ee ‘antage is the reduction in (E,/N) for a fixed bit area This reduction wlN) "ansmitted power and hence the hardware costs. f Da uanintevees The disadvantages of error control coding are Oa ST thd ee comes more “complex” due to implementation o} iS the 263 © scanned with OKEN Scanneron oem THO en Sang Let us now look in’o the significance of “redundancy”. a fi : the channel ¢ the transmitter systematically adds digits to the transmitted inessae® be These additional — digits carry “no information”, but make it pos’ ble for the channel decoder to detect and correct errors in the “information bearing digits”. This reduces the overall probability of error P, thereby achieving the desired goal. The additional digits which ong information are called “redundant digits” and the process of adding these digits is called “redundaney”, In the next section we shall consider a simple example of error control coding and show that there is great reduction in the probability of error P.. ‘There are several etror-correcting codes and these codes are classified under two basic categories namely “Block codes” and “convolutional codes”. The stinguishing feature for this classification is the absence of memory in the former case and its presence in the latter case, Another way of classifying codes is as “linear” or “non-linear”. A linear code differs from non-linear code by the property that any two code-words added using modulo-2 arithmatic, (which will be discussed later) produces a third code-word in the code. The codes used in practical applications are almost always linear codes. 5.3 EXAMPLE OF ERROR CONTROL CODING Figure 5.1 shows the complete block diagram of a digital communication system employing error control coding. The main functional blocks are the channel encoder, the channel decoder, modulator and demodulator and the noisy communication channel with a capacity C bits/sec. The source generates a message block {b,} at a rate of 1, bits/sec and feeds it to the channel encoder. The channel encoder, then, adds (n - k) number of redundant bits to these k-bit messages to form n-bit code-words. These (n - k) number of additional bits also called “check bits” do not carry any information but helps channel decoder to detect and correct errors. BIT RATE =, bits/sec (4) BITRATE =, bitssee INPUT MESSAGE CODED OUTPUT (4,} (b) CHANNEL ee aan BITRATE =1 = it BIT RATE =7, bits/sec] /-)ECODER is eee MODULATOR =} mbit Code-words — BLOCK OF k i Pomme? oe RaccRA Gis Sua caat MESSAGE BITS k MESSAGE oo” BITS Check Bits oo COMMUNICATION ‘CHANNEL, BLOCK OF k MESSAGE BITS n-bit Code-words Ie OUTPUT MESSAGE [TT DEMODULATOR i Da) DECODER loctimnene ania} Fi 1: Block diagram of communication system employing error control coding. an 2 © scanned with OKEN Scanner54 METHODS OF CONTROLLING ERRORS There are two different methods available for controlling errors in a communication system. (i) Forward-acting error corre receiver through attempts 10 correct noise-induced errors is called the forwar¢ correction method. Gi) Error Detection Method : In this method, the decoder examines the demodulator output, accepts the received sequence if it matches with a valid message sequence. If not, the decoder discards the received sequence and notifies the transmitter (through a reverse channel) regarding the error and requests for retransmission of the message till the correct sequence is received. Thus the decoder attempts fo detect errors but does not attempt to correct them. Error detection method yields a lower overall probability of error than error correction method. To illustrate this point, let us consider the previous example of transmitting the triplet ‘000’. If the decoder uses error detection method, then it would reject all other triplets except ‘000" and ‘111’. Now, an information bit will be incorrectly decoded at the receiver only when all the three received bits are in error. Thus P, = (q,)? = (8 x 10) = 5.12 x 10" which ismich lower than probability of error for error correction method. The disadvantages of error detection method are the requirements of reverse channel and slow down of the effective rate of data transmission. (This is because the transmitter has to wait for an acknowledgement from the receiver before transmitting next message). 55 TYPES OF ERRORS In digital communication systems, errors are caused by the noise present in the communication channel. Usually, two kinds of noise are encountered in communication siannels namely “Gaussian noise” and “Impulse noise”. Due to these, two\types of errors oeeur, \ A (@ Random Error : The transmission errors that occur due to the presence of white Caussian noise are referred to as “random errors”. Sources of Gaussian noise include thermal ‘td shot noise in the transmitting and receiving equipment, thermal noise in the channel and ‘dation picked up by the receiving antenna, _, {l) Burst Error: Impulse noise is characterized by long quiet intervals followed by thamplitude noise bursts. Examples of impulse noise are noise that arises due to lightning, thing transients, man-made noise etc. When such noise bursts occur, they affect more ‘ne symbol and the error caused is called “Burst Error”. ion method : The method of controlling errors at the ard acting error © scanned with OKEN ScannerInformation Tih 268 — 5.6 TYPES OF CODES 43] i As alpeudy mentioned in section, ).2, error control codes are divided into, two broad categories}namely “block codes” and. ‘convatutional codes”. iw 17940 @) Block Codes: Block code consists of (n~ k) number of check bits (redundant bits) being added to k number of information bits to form nv’ bit code-words. These (n—k) or check ails are “derived from k information Bits”. At the receiver. the check bits ar to detect and correct errors which may occur in the entire n-bit code-words. aT (i) Convolutional Codes : In this code, the check bits are continuously interleaved with inf rmation bits. These check bits will help to correct errors not only in Hi esicn block but also in other blocks as well. 5 5.7 LINEAR BLOCK CODES Giddedt In channel encoder, a block of *k” message bits is encoded into a block of *n’ bits/by — adding (n— k) number of check bits as shown in figure 5.2. Clearly n > k and such a code formed is called (n , k) block code. These (n ~ k) check bits are “derived” from k-message bits which will be shown in the next sec Ba CHANNEL plea tte saci) inna ents nk ENCODER (ni MESSAGE Ly MESSAGE —_|CHECK-BITS Fairs} I} kk te tt) oR CHECK-BITS] MESSAGE i (n-k) 1. k ——] Fig, 5.2 : Illustrating the formation of linear block codes A (nk) block code is said to be a “(n,, k) linear block code” if it satisties the cor given below : Let C, and C, be any two code-words (n-bits) belonging to a set of (n, k) block C, © C, {@ represents modulo-2 addition discussed in detail in next section] is al code-word belonging to the same set of (n, k) block code, then such K code i k) linear block code. ited cima A (n, k) linear block code is said to be “systematic” if the k-message bits ay at the “beginning” of the code-word or at the “end” of the code-word as de figure 5.2. utp of the row and column matrix yields =) oy + (0) (P,)) +» wt (1) (Pj) + ~ py) 0) +o “4(p,) () + a Pha (+ D=Py- “otra ble 52 ©) +O) +O) © “ = ws +P = Py hy Pith in matrix form, we have alues of i and j and hence, 404 (5.17) is Poy every V: | oe On coves re pondi ing elements on © scanned with OKEN ScannerInformation Theory and coding 274 vosue (5.19) Cy = Phnckhl 1 bove equation i ization of encoder mutator and ion results in the encoder for own in figure 5.3 consisting ‘The implementation of the al of modulo-2 adders. (n,k) linear block code, Such a rea of ak-bit shift register, a n-segment com circuit iss! (a —k) numbe: Kbit Shift Register Message sakes | ds | n-segment ‘commutator To channel Modulo-2 Adders Fig. 5.3 : Encoding circuit for (n, k) linear block codes ‘The entire data d, d,_; «dy dy is shifted into the k-bit shift register. The small circles Phy Patr-~ Pri" Phynnx ate either “open circuit” or “short circuit” dependi ither ‘0” “pe a : pending on either ‘0 or Ear example, if p,, = 0, then there is no connection from d, to the ere? adder and if Pi f i # aa there is oe When the message is shifted into the shift register, the: rae lulo-2 adders generate the | ‘check-bits? which are fed into the commutator segments along ecessively, a Pao ae pera Da vector bits will be transmitted through the Example 5.3 : For the systematic (6, 3) cod input of dy dd) is given by ) code of example 5.1 the code-vector C fora messiee © scanned with OKEN ScannerBW introsucnen —— 275 {C] = [d,,d,, Construct the Pete ie cedapnd ney Ned Solution The code-vector bits are given by Rae ln % =d,,¢,= 4, +d, c5=d, +d,,¢,=d, +d,. s = 3, we require a 3-bit shift register to move the message bits into it. We have —3=3 and hence we requir ; quire 3 modulo-2 adders and a 6 s entire encoding circuit is shown in figure 5.4 iat 2 Seance Message Input a wobenyar ™\ 3.bit Shift Register Commutator * yy To C channel Ce Cs Ce Fig. 5.4 : Encoding circuit for (6, 3) linear code of example 5.1 SYNDROME AND ERROR CORRECTION Let us suppose that C = (c, Cy, «». ¢,) be a valid code-vector transmitted over a noisy communication channel belonging to a (n, k) linear block code. Let R = (r, 1 «. ,) be the received vector. Due to noise in the channel rr... f, may be different from ¢, ¢... ¢. The is defined as the difference between *R’ and “ “error-vector” or “error pattern B” . BeRECeRE Gh Ol ch) Male as (5.20) Since subtraction is same as addition in modulo-2 arithmatic. «. The error-vector ‘e’ can be represented as a yector by see (5.21) E = (€, €) = &y) is clear that “E’ he error-vector is also an n-tuple where e, = 1 ifr, # ¢, and e, = ‘B? represent the errors caused by noise in the From equation (5.21), Dif r, =c,. The 1’s present in U channel. In equation (5.20), find E and then C, the receiv S defined as only ‘R’ and it does not know C and E. In order to the receiver knows i ow oding operation by determining an (n—k) vector er does the dec cai ieee © scanned with OKEN Scanner\ a _ Information Theory ay S = RH = (8, 8 ) n= The (n—k) vector S is called “error syndrome” of R. 7 From equation (5.20), R =C +E Using this in equation (5.22), we get § =(C+E)H" = CH™+EH" But CH! =0 from equation (5.18) Sahar pL were (5.24) ‘The receiver finds E from equation (5.24) as S and HT! are both known. Then from be found out, Note that the syndrome S equation (5.23) the transmitted code-vector ‘C’ can of the received vector will be zero if Ris a valid code-vector. When R#C, then S #0 and the receiver then detects and corrects the error. The following example clearly illustrates the method of sin 5.1, the received code-vector rred due to noise. gle error correction. Example 5.4 : Referring to the (6, 3) code of example R = [110010]. Detect and correct the single error that has occu Solution From example (5.1), we have 101 s(prm=|o 1 1/=PI 110 ol 1h 10 1 [P] = |0 ¥ -. From equation (5.14), eed ae a Ae 101100 @ % =/0 11010 c 110001 et 101 ro 110] 7P ae lee (Ay “t-enal=le | 010 001 From equation (5.22), the syndrome [S] is given by © scanned with OKEN ScannerRut =0110010 1 0 L 1 0 0 By using modulo-2 multiplication ang addition, the Syndrome is found to be S = [100] since teil S # 0, it represents an error. [The first syndrome bit s 1 is found from = I since the total number of 1's present is ‘odd’ If it is ‘even’ then the corresponding syndrome bit will be ‘0°. Consider sy = DOC.) OO1)@ (0.0) © (1.1) © (0.0) = °®1®0@0e1a@0 = 0 the number of 1's is 2 which is even number of 1's}. This syndrome vector S = (100) is present in the 4 row of HT matrix and hence the 4! bitin the received vector R counting from left is in error. «: The corrected code-vector is 110110 which is a valid transmitted code-vector as seen from table 5.3 corresponding to a message vector 110, SYNDROME CALCULATION CIRCUIT Let the received-vector R = (r, 1, «+ 1,). The syndrome vector $ is then given by equation (522) as am [S] = [s, 8, ......8, 4] =RH™ cys = [,5,.... Syod = Ot [Pa Pig, oe Pink Par Pap sss Paa-k Pur Pra 152.0) (oes | ia 0 Os it Olmert: 1 nghy using modulo-2 arithmatic, we get the syndrome bits as © scanned with OKEN Scanner«Information The, 278 — Ss TyPj + MyPar chess PY HI, . the Pie, 2th, Pip + MPa tee Pa + Tie F rd + (5.25) Bp Piyackat Pant ->++ +++ k Pink +h Equation (5.25) can be realized using the circuit shown in figure 93 whi “syndrome calculation circuit”. The received-vector bits are moved into a n-bit as shown. Here also the small circles p,,, Py), --- are either open circuit or short circuit depending on either ‘0’ or ‘1’. As soon as the received vector is shifted into the shift Tegister, the modulo-2 adders generate the syndrome bits S15 Spy Sy Knowing the syndrome Vector §, the error can be easily detected and corrected as shown previously. The following example illustrates the particular case of obtaining the syndrome calculation circuit, ch is calle Shift register ae F n-bit Shift | sulewdenes nf on Sea Fig. 5.5 : Syndrome calculation circuit for (n, Example 5.5: For the systematic (6, 3 R=[r, 5,5, 1,15 14]. Construct the comespor Solution For the (6 ,3) code, the matrix HT (ay? = 1) linear block code ) code of example 5.1, the received vector nding syndrome calculation circuit, is given by (refer example 5.4) moooen 10 o1 11 10 o1 00 © scanned with OKEN ScannerFrom equation (5.22), S = [s, s, s,] =R H™ S = (8, 8.83) = [r, 1) 6, 1,1, r6] f 0 1 1 0 0 0 1 om orFHOS 2 1 0 0 oe [@, +1, + 14), (0, +1, + 15), (t) +r, +1,)] .. The syndrome bits are Sian atar tt, 8, = r+ 1, +4, Satter) +4, <. The syndrome calculation circuit can be easil: 'Y constructed as shown in figure 5.6. Received Vector R —>| 8; 8; s, Fig. 5.6 : Syndrome calculation circuit for (6, 3) code of example 5.1 Rramnle § K+ Raein ovretamnntt BEL 2 (mari a @ scanned with OKEN ScannerCONVOLUTIONAL CODES ee 8.1 CONVOLUTIONAL CODES The main difference between block codes discussed in previous units and the convolutional codes (refer section 5.6 for definitions) is the following, In “block codes”, a block depends only on the block of n’ -digits generated by the encoder in a particular time-unit nput message digits within that time unit __ In “convolutional codes”, a block of ‘n’ code digits generated by the encoder in a time unit depends on not only the block of ‘k’ message digits within that time unit, but also on the preceding (m ~ 1) blocks of message digits (m > 1). Usually the values of ‘k’ and ‘n’ will be small. Like block codes, convolutional codes can be designed to either detect or correct errors. However, block codes are better suited for error detections and convolutional codes for error correction. Encoding of convolutional codes can be accomplished using simple shift registers and several practical procedures have been developed for decoding. ENCODER FOR CONVOLUTIONAL CODES : A convolutional encoder, shown in its general form in figure 8.1, takes sequences of message digits and generates sequences of code digits, In any time unit, a message block consisting of k digits is fed into the encoder and the encoder generates a code block consisting of ‘n’ code digits (k Code Blocks Fig. 8.1: General convolutional encoder 365 © scanned with OKEN ScannerInformation Theory and coq, ing _ ae 8.1: Anencoder for a(n, k, m) = (3, 1, 3) convolutional code is shown in figure 8, i t Explain the operation of the encoder and hence obtain the output of the encoder Commutstor | To channnel Fig. 8.2 : (3, 1, 3) convolutional encoder of example 8.1 Solution Let us look into the operation of the given encoder for a message input of 1 0110 (=4, 4, d, d,d,). 0 T, Pi eesti estan ety) Ft, Time ae ven Ee 0 | = Input message {d,} a fh o A! lore ‘opty! 00/0}! < Register contents F igds ‘001 0.00 '< Output | d,inflences these nine | d ‘— “ ‘output bitsi ' ' 1 ' i 1 Hl 1 ' influences these nine! 1 ‘i joutput bit rea i t 1 1 1 1 1 dyinfluences these nine 1 1 1 output bits Fig. 8.3 : Encoding operation of the convolutional encoder of figure 4.1 Operation : Let the shift register be cleared initially. The first data bit ‘d,’ is entered into the first flip-flop labelled D,. The commutator samples the modulo-2 adder outputs ¢, % and c,, Thus a single message bit yields three output bits. The next message bit ‘d,’ now enters D,, while the contents of D, which was ‘d,’ is shifted into D,. Then, again the © scanned with OKEN Scanner(367 WR, ociiona: Codes : i is repeated till the last message 1" outputs. This process is ; commutator samples abe a fia erate to first move d, into D, and then into D,, itis mi Teint ee emer ett atictalle Of the encoding operstion is shows in figure @ssumed that ‘0's a 8.3 along with the output of the encoder. f 1m convolutional encoder, the message stream continuously runs through the encoder unlike in block coding schemes where the message stream is first divided into long blocks and then encoded. Thus the convolutional encoder requires very little buffering and storage hardware, Ina general (n, k, m) convolutional encoder, the following notations are used. n = number of outputs = number of modulo-2 adders (normally) oe number of input bits entering at any time m = number of stages of shift register = number of flip-flops L = number of bits in the message sequence. Constraint length = mxn digits rate efficiency = k/n 8.2 ENCODING OF CONVO! APPROACH The concept of encoding of convolutional codes usin clearly understood through an example, Example 8.2 : Let us consider a (n, figure 8.4. LUTIONAL CODES USING TIME-DOMAIN ig time-domain approach can be k, m ~ 2, 1, 3) convolutional encoder as shown in Input @) Fig. 8.4 : Convolutional encoder (n, k, m — 2, 1, 3) of example 8.2 i i i Jutional encoder may be defined in terms ime- behaviour of a binary convol in germs ofa Brin ate responses”. The rate efficiency of the encoder of figure 8.4 is © scanned with OKEN Scanner“ Information Theory and Coding Therefore, we need to impulse responses to characterize its behaviour in time domain, {In general, we require “n” number of impulse Rey ali Let the sequence (2, 8, 2,0. y« ni denote the “impulse responses” also called “generator acl hia hareatplt path through “n” number of modulo-2 adders, In the encoder of figure 8.4, there are two modulo-2 adders labelled top-adder and bottom adder, Hence there will be two generator sequences. Let (dj, dy ... d,) represent the input message sequence that enters into the encoder, one bit at a time peartink, with d,. Then the encoder generates two output sequences, denoted by C"” and C®, defined by the discrete convolutional sums, given by CO = [deg C® = [d) #2? From definition of discrete convolution; we have w i) C= dees ; 5 (83) In the given encoder, j takes values 1 and 2 and i varies from 0 to m= 3. Let the message sequence be d, d, d,d,d,=10111. The output sequences are calculated as follows : For j = 1: From equation (8.3), av (8.1) (8.2) 5 a cM = 2s Bie where d,_,=0 forall output of the convolutional encoder is given by C= Cf €,% C7 CCPC, ont CGA in general for two modulo-2 adders. In the given encoder. c = 10000001 Cc =11011101 ©. The encoder output is | C = [11, 01, 00, 01, 01, 01, 00, 11] Tl METHOD (MATRIX METHOD) : The generator sequences g,‘” g,( gg, for the top adder and g, g, g,,. :© for the bottom adder, can be interlaced and arranged in a matrix form with the number of rows equal to the number of digits in the message sequence = L rows and number of columns = n (L+m). Such a matrix of order [L] x [n (L + m)] is called “generator matrix” of the convolutional encoder. @ In general, for a two modulo-2 adder convolutional encoder, the generator matrix G is | given by | 218 g2ep? B53 93? emai Bnet. 0-0 Oia oo | OOP rai heresies ently wees emeiO-s0% <0 0 0 CaO OO ND ee Fete Gori Basi... 0. 0 Smet Bnit J In the 2“ row, the number of ‘0’s is equal to the number of modulo-2 adders. Since the generator matrix G has n(L + m) number of columns, the encoder output will have n(L +™) number of bits given by C=dG “ean (8.6) In the given encoder of figure 8.4 d=10111 © scanned with OKEN ScannerSap ee ee an SI are wat e=1011 ez lta +. The generator matrix has rows and n(L. +m ) = 2 (5 +3) = 16 columns given by Pr Fore yyaniy 90 00 00 vo COE OTe Tey Gg 00 00 Ge OO OO roe tp 4h 00 00 00 00 00 oor oy 11 00 00 00 00 00 Ol tt ui The encoder output is given by equation (8 6) as C =aG " QO11yfIl Or 11 11 00 0000 00 00 11 Or 41 11 00 00 00 00 00 11 Ost 11 00-00 00 00 00 1 ott 11 00 00 00 00 00 m Or tt 11 C = [11, 01,00, 01, O1, O1, 00, 11] which is same as before. 3 ENCODING OF CONVOLUTION. DOMAIN APPROACH From the study of linear filter theory, we know that the convolution integral, which Scribes the linear filtering operation in the time domain, is replaced by the multiplication Fourier transforms in the frequency domain. Since a convolutional encoder is a LTI finite te machine, we may simplify computation of the adder outputs by applying an appropriate formation. Let the impulse response of each path in the encoder be replac: lynomial whose coefficients are represented by the respective elements of the impulse sponse. Thus for “j’” number of modulo-2 adders [where “j” varies from | to nj. We define genera‘or polynomial 89%) = 89 +B, 9X+BP RH + By OX” _ The corresponding output of each of the adders is then given by CO (x) = d(x) 9 (x) (8.8) After getting the polynomials at the output of each of the adders, the final encoder output Ynomial is obtained from s nes Cx) = CY (x2) + x CP (x2) + x7 CY (X8) desc XML CM. (xm AL CODES USING TRANSFORM- (8.9) le 8.3 : Obtain the output of the convolutional encoder of figure 8.4 using transform in approach. © scanned with OKEN ScannerInformation Theory and Coding 372 ng = 8.4 is given by The generator sequence for the top adder of figure 8.41 ) 2i)-= fa gyhgiMas= 011 given by equati ‘The generator polynomial corresponding to the top aa is given by equation (7) as w(x) = gi esx Hg Ot BE rx? a140¢r 40 Slt or sequence corresponding to the bottom adder of figure 8.4 is given by 2? = (2, 2,2 9,%2,01= 0 LEN The gen *. The Seat generator Seat is 2) 2 2) 424 9x9 g(x) = gM 49,2 x +g HEL X = lex¢x7 4x? . The top-adder output polynomial is given by equation (8.8) as C(x) = d(x) g&) We have, for the message d= 10 I 1 1, the message polynomial given by d(x) = 1 tx axtext CYR) = (ex 4x x4) (1+ x +x) Lex exe xt FR he 4 XP +x) CYR) = 14x? Similarly the bottom adder output polynomial is given by equation (8.8), with j = 2, C(x) = d(x) 2%) = (14x40 4x4) (14 x4x7 4x3) = l+xtx7+x) FO EXO” $x 4x txt 4x5 tO txtg x4 xo txt x54 x54 x7 CO(x) = Lex trendy xd ay? The final encoder output polynomial is given by e 3 oniiber af modulo-2 adders = 2, s y equation (8.9), with n = num CQ) = CY 2) +x C@ (2) (8.10) | We have C(x) = 14 x7 | 2 CO). = 14x “al } and 8 CRGR)I6 than acuecxtetadie 9 Gacy Lee Og hy 104 yt et © scanned with OKEN ScannerConvolutional Codes. “ ‘and (8.12) in (8.10), we get ee = VD RM Gg Cf x2 RO + RE + XI +X ? Wy xl x! Using Equations (g, Ih) x S Ltxe xt x7 + +X The code-w, rn oie Sy ord Corresponding to this polynomial is C = [11, 01, 00, 01, 01, 01, 00, 11] which is sa: S i ME as obtained using time-domain approach. @ scanned with OKEN Scanner