SK4082002A3 - Method for coding and decoding a data stream - Google Patents

Abstract

The method involves generating a binary code from source data streams for encoding and converting back into the source data streams for decoding. A source data stream is converted in a ternary data stream and a binary code with a maximum of four successive equal length discrete signal states is formed from the ternary data stream using state transitions.

Description

The present invention relates to a method of encoding and decoding data streams, wherein a binary code is generated from the source data streams when encoded and converted back to the source data streams when decoded.

BACKGROUND OF THE INVENTION

Currently, there are many encoding methods to generate RLL codes without DC voltage. Based on practical considerations, many of these methods work with constant-length code words (block coding). If the number of source symbols, respectively. many code words limited and countable, they are often used to encode and decode a table. There are known methods of encoding EFM and EFMPIus for writing data to optical or magnetic media (CD, DAT, DVD). Classical binary coding methods for generating signals without DC voltage are Manchester Code, Differential Manchester Code, 3-of-6, 8B10B (US 4,486,739) and CMI (Coded Mark Inversion) character inversion). These methods use two signal states (0 and 1 ”). In addition, non-linear coding methods are known, such as AMI code (AMI code), modified AMI code, HDB-3 (PCM30 transmission paths), 4B3T (national ISDN) code, MMS43 (ISDN Europe) or spectral lattice (Trel-lis- Shaping).

Existing methods for creating codes with a limited signal duration do not meet the requirements of special applications. Either the efficiency of the code is disadvantageous, the length of the code word does not match, the disparity exceeds the requirements, or the coding method is too costly for less demanding products. Especially preferred for radio transmission is low disparity coding in order to

31903 / T at the receiver of the radio signal drift the threshold levels in the demodulator (data slicer). Simple and space saving can be realized eg. code 3-of-6. With this code, 4-bit source symbols are converted to 6-bit code words. The disadvantage of 3-of-6 coding is its poor efficiency. Equals 4/6 = 66.7%; the loss in coding is thus 50%. For example, this may adversely affect the life and / or operating costs of battery-operated radio devices.

Coding methods other than 8B10B have a higher coding efficiency (E = 8/10 = 80%), but they can be implemented more costly because larger codeword tables and complex logic links are required.

In general, coding with tables over longer codeword lengths fails on the number of codewords that grows exponentially. Methods have therefore been proposed for enumeration of code words (see Kautz 1965, Pátrovics and Schouhammer Immink 1996). Counting decoding is performed by weighting the sums of the individual symbols of the received code words. The difficulty of this method lies in the determination of weight coefficients. Two sets of coefficients are often needed (see Tang and Bahl, 1970), whose determination represents an intellectual challenge in the field of combinatorics. With the increasing length of the code words, the number of weighting coefficients increases as well as the length of their expression in binary form. Therefore, in realizing the weighting coefficient counting method, the memory requirements with increasing length of the code words grow quadratic and therefore quickly encounter practical boundaries.

Non-binary coding methods such as AMI and others. cannot be used if the available hardware can only process two status signals.

31903 / T

SUMMARY OF THE INVENTION

SUMMARY OF THE INVENTION It is an object of the present invention to provide such a method of coding and decoding data streams that is easy to implement and not space-intensive, exhibits high coding efficiency, and generates a code that is suitable for radio transmission using technically simple radio devices.

In the coding and decoding method mentioned above, this task is essentially solved by transforming the source data stream into a ternary data stream (triad system numbers) for coding data streams, and generating binary code from this ternary data stream using state transitions. The binary code here has a maximum length (waveform) of four consecutive identical discrete signal states and a maximum disparity of two discrete deviations from the middle level of the two signal states or from the zero level; the binary code for decoding is transformed into a ternary data stream using appropriate state transitions and transformed into a source data stream.

In this way, a binary code is created from the input data stream (source data stream) without DC voltage and with a limited signal duration. By prior transformation to a ternary data stream, it is very easy to determine state transitions that allow the creation of binary code without specifying memory-intensive code words. E-encoding efficiency, i. the ratio of the length n of the source symbols in bits to length k of the code words, respectively. the encoded message in bits is log, (N) where N is the number of equally probable source symbols. As a result, the coding loss is designated as 1 /, i.e. the coding loss. relative extension of n-bit source message coded into k-bit code sequence

31903 / T

- _{1 =} ± _, nn minimizes log ₂ (N) E. This leads, in particular in the case of battery-operated radio data transmission apparatuses, to an improved use of the batteries and thus to an increase in the operability of the apparatuses and / or a reduction in costs.

Under the boundary conditions of the maximum disparity of the code to be generated with a maximum of Dmax = ± 2 and a wave length of maximum RLmax = 4, an optimal marginal efficiency of the E. _x coding method can be determined. Lies around E _x ~ 79.25%. Theoretically achievable minimum coding loss V _x at E n will be about 26.2%.

According to the present invention, the state transitions of the individual numbers (0, 1, 2) of the ternary transcription of the data stream are always converted into two binary digits of the binary code (00, ΟΙ, IO or II).

To achieve a clear transformation between thematic and binary

In the preferred version of the present method, the definition of state transitions depends on the normal disparity of the data stream. The normal disparity results from the overall parity of the binary code word taken into account so far. If the requirement is the absence of DC voltage, e.g. if the binary state of signal 0 generates a disparity of D = -1 and the binary state of signal 1 generates a disparity of D = +1, one binary codeword 00I has the sum of the total disparity D = -1. By adding binary combination II to binary code word II00I, the total disparity (normal disparity) is changed to D = +1. If now the state transition T2 e.g. depends on the common disparity D such that at D <0 T2 the ternary number 2 is converted to binary combination II and at D> 0 it is converted to binary combination 00, by means of state transitions TO, T1 and T2 a unique transformation between ternary states 0 , 1, 2 and binary states O1, I0, 00, respectively. H.

31903 / T

In order to create a defined initial state, the trigger bit or the disparity of the previous bit sequence can be used to create a trigger disparity.

Especially when computers are used, the source data stream is available as binary source code. In this case, according to the present invention, the binary source data stream in blocks of predefined length (block length) is transformed into a dark data stream. Preferred block lengths of the binary source code are 11, 19 or 84 bits, because at such lengths the coding efficiency is very close to the achievable optimum.

In accordance with the invention, before the data stream transformed into the binary code is preceded by a SYNC sequence, the recognition of the beginning of a new, stronger radio message level, which overlaps the weaker message level, is made possible. In cases where messages overlap, which can happen in unregulated radio traffic, this creates the possibility of detecting a collision of radio messages. With sufficiently large differences in reception levels, the stronger of the two radio messages may be enforced.

For this purpose, the SYNC sequence comprises, in particular, a synchronization sequence with a sequence of alternating signal levels and a trigger sequence with a waveform (signal) of successive same signal states greater than the maximum waveform length. Due to the greater signal length, the trigger sequence in the data stream can be more easily identified.

In order to be able to easily connect the coded data stream to the trigger sequence, the trigger sequence according to the present invention ends with a defined disparity, in particular zero disparity, or simply positive or negative disparity. The trigger state is thus precisely defined.

31903 / T

BRIEF DESCRIPTION OF THE DRAWINGS

Next, the preferred coding and decoding method is described in more detail with reference to the figures. Other advantages and features of the invention will emerge from this description and / or figures, irrespective of how they are formulated collectively in the claims or their interrelations.

The images show:

Fig. 1 shows a table summarizing the number of code words C _D (2n) with a signal length limitation and a balanced bit ratio of 0/1 at different disparities;

Fig. 2 shows a binary code construction scheme for a 16-bit codeword with a signal length limitation RL _M ax = 4 and a maximum disparity D _M ax = +/- 2;

Fig. 3 shows the state machine for generating the binary code of FIG. 2;

Fig. 4a, b construct a SYNC sequence;

Fig. 5 shows a state machine for encoding the reception of the binary code of FIG. 2;

Fig. 6 shows the net excess code of the RLL code as a function of the codeword length;

Fig. 7 shows the net coding efficiency of the RLL code as a function of the code word length; and FIG. 8 percent deviation from the absence of the DC voltage of the RLL code as a function of the code word length.

31903 / T

Ί

DETAILED DESCRIPTION OF THE INVENTION

Further described by way of example is a method of coding and decoding data streams, which is particularly suitable for generating a signal flow without DC voltage and with a limited signal duration. For example, a DC-free signal flow may be needed to modulate the carrier high frequency during radio transmission, or to transmit data through lines that simultaneously feed the supply current. The focus of the method of the present invention is to provide efficient radiotelegrams for transmitting measurement data.

An elementary condition that binary codes do not contain a DC voltage is that the duration of the signal state in aggregate equals the duration of the complementary state of the signal. For data transmission by a discrete, constant step cycle in a binary code, this indicates that the number of n 0-bits must be equal to the number of n l-bits. For example, let the code words be 2n. Then the number of C (2n) code words with a balanced bit rate of 0/1 without limiting the length of the waveform will be:

C (2n) = ( ² ), where not N

If a coding is introduced by limiting the number of consecutive equal signal states (signal length limitation, RLL code) by deviating the signal flow from the middle level of the two signal states, respectively. from zero level (disparity D = D _M ax), then the number of C _D (2n) code words with 2n bits length will be

C _D (2n) = f2n ^A n-1 + Σ (-1 / (2n ⁵ l n J i = 1 <n- (D +) i>

31903 / T where

for (m <0) in (m> n)

Fig. 1 contains a table with an overview of the number of code words C _D (2n) of different length for different disparities.

For the disparities D = 1 and D = 2, C _D (2n) can be readily ascertained. Especially the ratio

C ₂ (2n) / C ₂ (2n-2) is constant, ie the systematic construction of the code is evident. Case D = 1 is unattractive because of the coding efficiency (E = 50%, as in the Manchester code). In case of D = 2 the optimal boundary efficiency E /

E '= lim log; (2.3)

2 log log, (2) + (--l) log, (3) lim. - 2n £ = lim lim n - n log, (3)

2n + lim n-> oo 10g; (3) 2n log; (3) 2 ä 0,79248125 «79,25% coding loss at borderline efficiency is:

V _n = - - 1 * 0.26185951 26 26.2% E _x

31903 / T

A detailed description of the construction and creation of a binary code with a signal length limitation RLmax = 4 and a maximum disparity Dmax = 2 is given below. Figure 2 schematically shows the possibilities of generating such a binary code, the total code word disparity being shown in bits.

The starting point before writing the first bit is the disparity D = 0. is l-bit, the disparity is D = +1. If the second bit is also l-bit, the disparity will have a maximum value of D = +2. In order not to exceed the limits of maximum disparity, only the O-bit can be entered as the next one. In order to reach the lower limit of disparity D = -2, a maximum of 4 O-bits can now be written. It can be seen from the diagram that the code created contains at most 4 identical single symbols 0 or I, and thus has a maximum waveform (signal) RLmax = 4. Disparity ranges from +2 to -2 and at the end is D = 0; therefore the code does not contain DC voltage.

Thus, a state machine can be created that is capable of converting input data streams of any length into output data streams (codes) without 1 DC voltage, with the required properties Dmax = 2 and RLmax = 4. Such a state machine is shown in FIG. Third

The starting point for the state machine is the input data flow (not shown). Depending on the information it contains, this flow can have any value and can be transformed into binary form. For example, the following binary input data can be assumed:

Input data, binary 1111111111111111111 (19 bits, value: 2 ¹⁹ · ¹⁾ .

This input data will be transformed into a ternary number (i.e., a number in the triple or base 3 system) in the first (also not shown) intermediate step and will now look like this:

Intermediate, ternary 222122012001.

31903 / T

Without limiting generality, the state machine is now in a state of positive or negative disparity (D + or D-). The D + or D states can first be reached by a defined sequence of SYNC or (starting from the trigger state S) by the state transitions TO or T1 with a separate trigger data bit. Depending on whether this first bit is 0 or I, D + or D- states are reached. In the start-up state, the state transitions are defined as follows:

T0 (S): write bit 0 a in binary output code

T1 (S): write bit I in the binary output code

In this example, make the first bit of the input code I and use the state transition T1. It writes l-bit in the output code and creates a code word with positive disparity. The state machine is therefore in a state of positive D + disparity.

The state machine now takes the first digit (in the example from the right) of the current intermediate step and assigns it to one of the state transitions

T0 (D +): write in binary output code sequence of bits Ol,

T1 (D +): Write the IO or bit sequence in the binary output code

T2 (D +): Write a sequence of bits 00 in the binary output code, with the number 0 assigned to the state transition TO, the number 1 to the state transition T1 and the number 2 to the state transition T2.

In the example, the first ternary digit is from the right 1. Therefore, the state transition T1 (D +) adds a sequence of IO bits in the output code. The entire word in the output code is then I0I and also has a simple positive D + disparity. For the other two ternary digits 0, the state transition T0 (D +) in turn adds a sequence of bits Ol to the output word, so that the entire code word OlOIIO1, which also has a positive disparity, is produced. For the next ternary number 2, the state transition T2 (D +) is now used, and it adds a sequence of bits 00 to the total codeword OOO1010II. This word will now have a negative D- disparity.

31903 / T

For states of negative disparity D-, state transitions are now defined as follows:

TO (D-): write the sequence of bits Ol in the binary output code,

T1 (D-): write a sequence of bits I0 or binary in the binary output code

T2 (D-): write the sequence of bits II in the binary output code.

Therefore, for the next ternary number 1, the state transition T1 (D-) is used. By correspondingly using state transitions, we get the following output code, using the state transition {} (D +) -> 0 or {} (D -) -> l 'at the end of the word to reach the final state E without DC voltage:

Output code, binary 0II00III000II0II0000I0II0I

In the previous example, it was assumed that all the input data to be transformed consist of a sequence of 19 bits and one additional trigger bit. In practice, however, the input data is usually longer. Transforming large numbers from a binary system (binary system) to a thematic system (triad system) is not a problem, but it requires an increasing number of computational operations. Therefore, splitting (= chopping) the input data into blocks of a fixed length is a certain convenience. In this case, the lengths of the blocks are selected so that there is as little as possible in the respective ternary transcription of such chopping, i. to make good use of the area of transcript numbers for a given number of locations.

When coding in blocks, it is possible to switch from the start state S to the operating states D- and D-state respectively. D + in that the first bit of the output stream is selected as fixed or random, or the first bit of the input stream is selected for the TO and TK transits respectively. T1.

At the beginning of the transformation, i. Thus, for the first block to be transformed, the D + or D- disparity is specified. All subsequent input data as well as outside the boundaries of the blocks can then be made

31903 / T move only within the limits of both these D + and D- states. Only at the end of the encoding, i. after encoding the last block of input data, the disparity is removed by appending the last bit ensuring complete absence of DC voltage. In this case, the states S and E of the state machine are irrelevant to the coding of a single block of data, with the result that the complete absence of DC voltage is not reached at each end of the block but only at the end of the message.

If the input data stream is binary and consists of b bits, then 1 bit can be used directly to create a trigger disparity D + resp. D-. If we consider the remaining (b - 7) bits to be a binary number, then this number can be transformed from a binary to a ternary system (base 3 numbers) by changing the base of the number system. The individual digits 0, 1 and 2 of the ternary notation can now serve directly as input conditions TO, T1 and T2 of the state transitions. In order to limit the number of computational operations, it is useful to transform the remaining (b - 7) bits of the dual write in blocks. Suitable block lengths are those in which the n-th power of 2 is only slightly less than the n-th (n is an integer) power of 3, i.e.:

2 3 3 ^m A2 <3 ^m where m, n & N, in particular

2 ¹¹ = 2048 »2187 = 3 ⁷ resp.

2 ¹⁹ = 524288 = 531441 = 3 ¹² , respectively.

2 = 19342813113834066795298816 ⁸⁴ '19383245667680019896796723 = 3 ^53rd

The effectiveness of the code is:

in the case of (2 ¹¹ θ 3 ⁷ ) E _11/7 = 11 / (2x7) «78,571%, in the case of (2 ^1θ θ 3 ¹² ) E _{19 / l2} = 19 / (2x12)« 79,167% and in the case of (2 ⁸⁴ θ 3 ⁵³⁾ E _84/5 3 = 84 / (2x53) "79.245%.

31903 / T

Therefore, it is advantageous, for example, to divide the data stream (to be coded) into 11-bit words and transform these in blocks into a 7-digit ternary notation. In another variation, it is also possible to divide the binary input data into 19-bit words, which are then transformed into a 12-digit ternary notation. In the third variant, it is proposed to divide the binary input data into 84-bit words that are transformed into a 53-digit ternary notation. In all three cases, the ternary digits {0, 1, 2} are converted to four possible output states (00, ΟΙ, ΙΟ, II) without intermediate step, depending on normal D disparity (ie D + or D- state).

However, it is also possible to apply the entire state machine to each individual block. It would then be advantageous to select bit blocks of data with a length of 11 + 1, 19 + 1, respectively. 84 + 1 bits, the additional bit being output from the start state S. From (11 + 1) the data bit block is coded (1 + 2 * 7 + 1) = 16-bit word in RLL, z (19 + 1) ) bits will be (1 + 2 * 12 + 1) = 26 bits, and (84 + 1) bits will be (1 + 2 * 53 + 1) = 108 bits. Thus, after each block of data, the absence of DC voltage is obtained at the cost of longer code words and at the cost of impaired coding efficiency (E _{11 + 1} = 12/16 <E _1V7 = 11 / (2x7); E _{19 + 1} = 20/26 <E _19/12 = 19 / (2x12), and E ₈₄₊ = 85/108 <E _{8 4/53} = 84 / (2x53).

State machines with equivalent properties can be created by reversing the input conditions TO, T1 and T2 in state transitions.

Specific boundary conditions apply to the field of use of radio transmission. In systems with multiple equally authorized transmitters in which the transmission of messages by radio is not coordinated with each other, radio message collisions may occur. Such systems are therefore required to be able to recognize conflict-disturbed messages. In addition, messages with a higher level of income overlapped with messages with a lower level of income should be capable of being received with sufficient interference. Especially during the reception of one radio message, another, later incoming radio message with a stronger reception level should result in normal reception being recognized as disturbed and interrupted and a new, overlapping stronger message can be received. To this end, before

The 31903 / T own data is preceded by a SYNC sequence that is used in the radio receiver to adjust the receiver to the signal strength of the received message and to obtain a step cycle for the portion of the information that follows following the binary signal sequence.

The SYNC sequence consists of a clock sequence and a start sequence. The clock sequence is a step frequency defined as a sequence of alternating signal levels. The clock sequence must be selected so long that the receiver reaches a steady state at the end of the clock sequence. The clock sequence is followed by the start sequence. It is defined as a sequence of signals that can be clearly distinguished from the RLL data stream. Since in the described coding method there are no reserved code words with a permissible waveform length RLmax = 4, the runtime length of the trigger sequence is greater than that of the code RLL, t. no. RLmax> 4. SYNC sequence design options are shown in Figures 4a and 4b.

The distinguishability of SYNC sequences from the RLL data stream is given by five consecutive zeros (Fig. 4a), respectively. sequence of five ones (Fig. 4b). The disparity at the end of the SYNC sequence is zero so that the state-coded data stream can be seamlessly connected.

The reception decoding state machine is shown in FIG. 5. It is used to translate the binary code back into the source code. The function principle can be compared with the function principle of the state machine shown in FIG. 3 and used to generate a binary code, wherein the state machine can also recognize the SYNC sequence.

Beginning with the trigger state, the scan operators 0 and I are first monitored for the alternation of bits at the clock frequency, which periodically alternates states between BO (per bit 0) and B1 (per bit 1). As soon as the regular bit rotation is interrupted by two consecutive bits, the TS step tests whether a defined trigger sequence is available. Until a valid trigger sequence is detected in step TS, the prescribed steps are repeated.

31903 / T

After recognition of the trigger sequence, the data recognition is first entered into a DO state with a disparity of D = 0, which consists of selecting a binary data code. In order to recognize the initial state of disparity (D- or D +), it is determined by the operators 0 and 10 respectively. First, the first bit used as the trigger bit to determine the trigger parity. This enters a state of negative (D-) or positive (D +) disparity. Then the other two bits are read as a sequence of bits and translated into ternary digits using state transitions. These status transitions are:

0l (D- / D +) write ternary digit TO = 0,

IO (D- / D +) write the ternary digit T1 = 1, II (D-) write the ternary digit T2 = 2 ,.

00 (D +) write the ternary digit T2 = 2.

Furthermore, the OO (D-) state transitions, respectively, are defined. II (H +). If these state transitions have to be used based on a sequence of bits entered in the binary code, a state with a disparity greater than the specified maximum disparity Dmax = 2 has occurred. This indicates that the data word is disturbed (eg, by overlap). Therefore, as a result of these state transitions, the end of the message is written as initial data, and the state machine enters a TS state to check if a new trigger sequence is available.

If there is only one bit in the data stream, this indicates the end of the data to be decoded. In this case, the state automaton is converted by means of operators I and II respectively. Finally, the ternary number is transformed back into the source code.

In order to simplify the feasibility, a variant is proposed in which the SYNC sequence simply ends in a positive or negative sequence. the negative disparity and the coding of the data stream begins in the D- and D-state respectively. D +. This saves the trigger state S while implementing the encoding method and the DO state when decoding,

31903 / T

Figures 6, 7 and 8 show: net coding excess, net coding efficiency, and percentage deviation from the absence of DC voltage RLL as a function of codeword length.

The present coding and decoding method can be implemented in a simple and space-saving manner. This is particularly important for products in the low cost area. This method creates binary codes of any length from the ternary data stream, which can be easily decoded back into the ternary data stream. At the boundary conditions RLLmax = 4 and D _M ax = 2, the coding efficiency near the theoretical optimum is 79.248%. Moreover, this method suffices without memory-intensive code word tables and the length of the code words to be created is in principle unlimited. Using a constant ratio of C ₂ (2n) / C ₂ (2n-2), the codeword can be systematically constructed at different codeword lengths.

Claims (9)

A method of encoding and decoding data streams, wherein binary code is generated for encoding from source data streams, and for decoding, the binary code is back-converted to source data streams, characterized in that for encoding data streams, the source data stream is transformed into the binary code has a maximum waveform length of four consecutive discrete signal states and a maximum disparity of two discrete deviations from the middle level of the two signal states or from the zero level, and that the binary code for decoding by appropriate state transitions, it is transformed into a real data stream and transformed into a source data stream. Method according to claim 1, characterized in that the state transitions associate individual digits of the ternary notation of the data stream in each case to two binary locations of the binary code. Method according to claim 1 or 2, characterized in that the definition of the state transitions depends on the normal data flow disparity. Method according to claim 3, characterized in that the trigger bit or the disparity of the previous bit sequence is used to generate the trigger disparity. Method according to one of claims 1 to 4, characterized in that the source data stream is available as binary source code and transforms into a regular data stream in blocks of a specified length. The method of claim 5, wherein the length of the binary source code block is 11, 19 or 84. 31903 / T Method according to any one of the preceding claims, characterized in that a SYNC sequence is preceded before the data stream transformed into the binary code. The method of claim 7, wherein the SYNC sequence has a clock sequence with a sequence of alternating signal levels and a trigger sequence with a sequence of successive equal signal states greater than the maximum data stream length. Method according to claim 7 or 8, characterized in that the trigger sequence results in a defined disparity, in particular a zero disparity or simply a positive or negative disparity.