FR2967320A1 - METHOD AND DEVICE FOR DECODING DATA RECEIVED IN FRAMES - Google Patents

Abstract

Method for decoding data received in frames, * data received in frames having differential modulation and convolutional coding. The method applies turbo demodulation (17) and deinterleaving (12, 13) as well as convolutional decoding (20). The deinterleaving (12, 13) is done in the time before the turbo demodulation (17), the turbo demodulation (17) and the convolutional decoding (20) being done by iteration processing. The convolution decoding (20) applies at least in part a flexible output Viterbi algorithm.

Description

FIELD OF THE INVENTION The present invention relates to a method and a device for decoding data received in frames. In digital radio transmission systems such as DAB (Digital Audio Broadcasting) standard, a multi-carrier method is used and each carrier is modulated by differential four-phase manipulation (4-DPSK). Phase shift keying (PSK) is a phase modulation of digital signals. According to this method, the signal has a constant frequency and a constant amplitude. The phase of the carrier signal varies abruptly with the rhythm of the digital modulation signal. For DAB transmission, several audio streams are combined with pure data services, also possible, to have a higher data rate set. The multiplex signal obtained is modulated in OFDM mode (modulation of orthogonal frequency division signals). The OFDM mode is a modulation method using several orthogonal carrier signals for the transmission of digital data. Each carrier is modulated separately. In addition, according to the DAB standard, to separate the grouped fields, a frequency interleaver and a time interleaver are used. To correct the errors, a dotted convolution code is used. In addition, according to the DAB ETSI 300.401 standard, the demodulation is carried out in the receiver by so-called soft decision demodulation and the convolution decoding is done using a Viterbi decoder. The use of a convolution code allows self-correction of errors. Thus, by introducing a redundancy, one arrives at a higher protection against the errors of transmission. The mathematical convolution method thus distributes the information content of the various useful data positions into several code word privileges. The convolution decoding in the receiver using a Viterbi decoder is based on the processing of the different positions of a codeword with determined probabilities. The Viterbi algorithm is a dynamic programming algorithm for determining the most likely sequence of hidden state. The decoder

Viterbi can process binary input sequences as well as continuous input sequences. STATE OF THE ART The document "C. Berrou, A. Gelavieux, P. Thitimajshi: Near Shannon Limit Error-correcting Coding and Decoding Proceedings of the IEEE International Conference on Communications, 1993: ICC 93, Geneva" describes the principle of decoding turbo convolution code nested in parallel. The document "Dr. Ing. Thomas May: Differenzielle Modulation und Kanalkodierung in breitbandigen OFDMFunkübertragungssystemen; Dissertation; TU Hamburg-Harburg "describes the combination of a differential modulation and a simple convolution code, which formally corresponds to a concatenated convolution code and thus makes it possible to apply the principle of turbo decoding. EP 1 284 548 A2 discloses a method and a device for association by an index of data received in frames. This allows a dynamic association of data to a channel in a memory. It is an object of the present invention to develop a method and apparatus for decoding received data in frames such as, for example, data received by digital radio signals by improving these means so that for a particular the same signal / noise ratio, there will be a lower error rate or that for a reduced ratio, it remains below the maximum error rate. DESCRIPTION AND ADVANTAGES OF THE INVENTION For this purpose, the subject of the invention is a decoding method of the type defined above, characterized in that at least one turbo demodulation and one deinterleaving and one convolution decoding are applied. , * the deinterleaving being done in time before the turbo demodulation, * the turbo demodulation and the convolution decoding being done according to an iteration processing, 3 * the convolutional decoding applying at least in part a flexible output Viterbi algorithm. Turbo demodulation here denotes a decoding method with several decoders in parallel or in series. For this, statistical information is exchanged between the decoders and the decoding process is done by iteration. The term "deinterlacing" refers to "descrambling". For this, the data that has been scrambled on the transmitter side to reduce burst errors, are again put back in their correct order. The additional steps of decoding the differential modulation and convolutional coding data are also provided according to the state of the art. It is for example a Fourier transform such as fast Fourier transform FFI 'or discrete Fourier transform DFT to transform each symbol in the corresponding frequency range. Such a Fourier transformation is for example applied as one of the first essential steps of the data decoding method. According to the invention, the time organization of turbo demodulation, deinterleaving and convolution decoding has been transformed to effect de-interlacing prior to turbo demodulation. Deinterlacing may be the combination of frequency deinterleaving and time deinterleaving. The decoding method according to the invention is further characterized in that the turbo demodulation and the convolution decoding are performed in an iterative processing loop. Convolution decoding is then based at least in part on the flexible output Viterbi algorithm (SOVA algorithm). In addition to the decoding of the different bits, the SOVA algorithm also generates information about reliability. The SOVA algorithm is a modified Viterbi algorithm. In the SOVA algorithm, for each information bit, the probability of error is calculated and the soft output is used to provide the probability for the decision bit for the downstream soft decision process. Thanks to this special chronological organization of the process steps, the turbo decoding principle can also be used for differential modulated or coded data.

4 convolution also in systems such as digital radio broadcasting systems without having to have a larger memory or computation capacity. Thus, for the same signal-to-noise ratio, there will be a lower error rate. In addition, for a lower signal-to-noise ratio, for example 2.5 dB, one will go below the maximum error rate. This results in a gain in the AWGN channel (Gaussian white noise) of up to 2.5 dB. The abbreviation AWGN denotes Gaussian white noise which is a white noise whose signal amplitude has a Gaussian distribution and whose noise power spectral density is constant. For the reconstruction of the multiplex signal, it is preferable to associate the data, in particular to associate the real part and the imaginary part of a bit by means of indexes as well as an interleaving memory. This concept of index makes it possible to place the corresponding data, for example the real part of a bit, and its imaginary part being in different places and even in different codewords, to associate them again in a different way. guarantee. This solution has the advantage that the data can remain at their location in the memory and the association is done only by an index which is in an array in the memory. Preferably, the method according to the invention is applied to a reception channel of a digital radio receiver operating according to a digital radio system such as DAB, DAB + and DMB. The method may further be used as a transmission method applying differential modulation such as DAB, DAB + and DMB. The method can be applied to all the services of the transmission methods provided by the standards. Preferably, soft outputs are generated by the SOVA algorithm only for non-spotted locations in the convolutional code. Since the demodulator can not use the dashed locations of the flexible outputs generated by the iteration processing loop of the convolution decoder, it is thus possible to reduce the means implemented by virtue of this preferential measurement. Soft outputs indicate the probability that a bit is "0" or 35 "1". On the other hand, the rigid outputs are decided bits. Each rigid output 2967320 is thus associated with a flexible output. The dashed positions are the bit positions of a codeword that has been removed on the transmitter side by a "dotted line". With the dotted emitter side, the code word lengths are precisely designed over a defined frame length for subsequent transmission of the data. On the receiving side, there is provision for a dotted line. In addition, preferably, in the flexible outputs of the demodulator, the dashed positions are replaced by zero positions. Thus, the dotted lines of the flexible outputs of the demodulator are absolutely not removed, but a zero is introduced at each dotted position by the convolution decoder or the SOVA algorithm without using the flexible output values of the memory. This allows memory savings for the dashed positions until final decoding of the convolution code. In a preferential manner, it is only after the last iteration that the dotted lines are eliminated in the output of the rigid outputs. In addition, the values of the soft outputs are preferably used as stop criteria for the iteration processing loop. As soon as the value of a soft output exceeds a predefined limit value, the result of the decoding is considered sufficiently reliable and the iteration processing is stopped. Preferably, after a maximum number of iterations initially defined, if after this maximum number of iterations, there are still very small amounts of soft outputs, this indicates that a decoding called "d" has been used. 'erasing' an external code or an error confined. These means are necessary to avoid possibly damaging the reconstructed signal by error positions in particularly sensitive information. Since differential modulation in the direction of time makes it possible to demodulate OFDM subcarriers independently of one another, it is possible to further reduce the iteration process again by iteration only the subcarriers whose convolution decoder provides flexible outputs with very small amplitudes as this implies a particularly uncertain evaluation. This saves a new bit demodulation that has already been decided for sure. The decoding device according to the invention of data received in frames comprises a demodulator and a convolution decoder. The demodulator is used to demodulate the differential modulation data and according to the invention, it is implemented as a turbo demodulator. In addition, the decoding device comprises at least one deinterleaving formed of two separate interleavers, a time deinterleaver and a frequency deinterleaver. The convolution decoder according to the invention is an SOVA decoder and forms with the demodulator, an iteration processing loop for the data received to perform the demodulation. Advantageously, the device comprises an interleaving memory for reconfiguring the multiplex signal. This deinterleaving memory makes it possible to use an index concept to associate, for example, the real part of a bit and its imaginary part at different locations and even in different codewords. Drawings The present invention will be described in more detail below with the aid of examples of data decoding method shown schematically in the accompanying drawings in which: - Figure 1 is a block diagram of a chain of 1 is a block diagram of the turbo decoding principle for DPSK and the known convolution code according to the state of the art "Dr. Ing. Thomas May: Differenzielle Modulation und Kanalkodierung in Breitbandigen OFDM-Funkübertragungssystemen »- Figure 3 is a block diagram of the method and the device according to the invention, of decoding data received in frames, - Figure 4 shows another block diagram of the method and the device according to the invention for decoding data received in frames. DESCRIPTION OF EMBODIMENTS OF THE INVENTION FIG. 1 shows a characteristic reception string of a DAB system such as that known for example according to the ETSI 300.401 standard for digital radio systems. The receiving s-chain processes the received data in frames by OFDM transmission, with differential data modulation and convolutional coding, by a Fourier transform to transform them so that in return each symbol is transformed back into the corresponding frequency range. By consecutive differential demodulation, the states of the complex signals are recovered. As according to the DAB standard applied in the transmitter, the data has been frequency and time interleaved to spread out the beam defects, the demodulated data in the receiver is subsequently subjected to the demodulation 11 in the compound de-interleaver. 1s of a frequency deinterleaver 12 and a time deinterleaver 13. Next, the characteristic DAB reception chain provides a dotted line deletion 14 and soft decision demodulation by an appropriate Viterbi decoder 15. Figure 2 shows the principle of turbo decoding for differential modulated PSK signals (DPSK) and convolutional coding. The document "Dr. Ing. Thomas May: Differenzielle Modulation und Kanalkodierung in Breitbandigen OFDM-Funkübertra gungssystemen "indicates, for example, that the combination of a differential modulation and a simple convolution code formally corresponds to a convolution code with concatenation and thus makes it possible to apply the turbo decoding principle. For this purpose, a SOVA demodulator 17, a convolution decoder 20, a deinterleaver or a frequency deinterleaver 12 and a time deinterleaver 13 as well as an interleaver 19 are combined in an iteration processing loop. However, the practical application of this principle to a DAB system is very complicated. In the assembly of FIG. 2, the interleaver in transmission mode DAB 1 would for example pass over 288 OFDM symbols. In combination with the frequency deinterleaver, this would require a large memory capacity and a large computing power. To avoid such a need for large memory capacity and high computing power for the turbo decoding of FIG. 2, the deinterlacing is carried out using the frequency deinterleaver 12 and the deinterleaver in time 13 before the demodulation. This is possible because the frequency interleaving requirement relates to the position of the carrier and will not be changed by the demodulation. The time deinterlace prescription is repeated every 16 bits. The modulated bits on the real part and the imaginary part of a subcarrier are separated by a configuration. These bits are put in code word positions that differ in the number of carriers. As in the DAB system, in all modes, the number of carriers is a multiple of 16, the same ls interleaving time is applied to the bits of the real part and those of the imaginary part. This makes it possible to put the time deinterleaver 13 before the demodulation. Such an arrangement in which the demodulator 17 is downstream of the Fourier transform 10, the frequency deinterleaver 12 and the time deinterleaver 13 is shown in FIG. 3. The iteration loop for turbo demodulation as shown also FIG. 3 is carried out by the SOVA demodulation functional blocks 17 with an SOVA convolution decoder 20 as well as the dotted suppression device 14 and a dashed dashing unit for the dots 21 of the flexible outputs 22. To simplify the repetition dashed lines 21 and 25, the dotted line deletion 14, in the iterative loop, the demodulated flexible outputs 22 are completed in their shape only once to the deletion of the dashed lines 14, then replacing the dots at the locations with zeros. In subsequent demodulation steps, the flexible outputs 22 are directly overshot at the correct locations. FIG. 4 shows a block diagram based on the dis-position of FIG. 3 and whose dotted line deletion 14 is after the convolution decoder SOVA 20. Thus, in the iteration loop, the suppression of dashed lines is avoided. and the dotted lines 21. This makes it possible to reduce the computing power required. In this case, memory can also be saved for the dashed locations until the convolution code is finally decoded. In this case, the dotted lines of the flexible outputs 22 of the demodulator 17 are not removed. The convolution decoder SOVA 20 always replaces the dashed po- sitions with a zero without accessing the memory of the soft output values. It is only after the last iteration that we remove the dots 14 to emit the rigid outputs 23. io io NOMENCLATURE

10 Fast Fourier Transformation 11 Differential demodulation 12 Frequency deinterlacing 13 Time deinterlacing 14 Dotted line deletion 15 Viterbi decoder 16 Memory io 17 SOVA demodulation 18 Deinterlacing 19 Interleaving 20 Convolution decoder 21 Dotted / dotted set-up 15 22 Flexible output 23 Rigid output 20 25 30

Claims (1)

CLAIMS1 »Method for decoding data received in frames, * data received in the frames having differential modulation and convolution coding, characterized in that - at least one turbo demodulation (17) and one de-centralization are applied lacement (12, 13) and convolutional decoding (20), the deinterleaving (12, 13) being done in time, prior to turbo demodulation (17), turbo demodulation (17), and the convolutional decoding (20) being done by iteration processing, the convolutional decoding (20) applying at least in part a flexible output Viterbi algorithm. 2. The method as claimed in claim 1, characterized by a dynamic association of the data by indexes, the association serving in particular to associate the real part and the imaginary part of a bit. 3. The method according to claim 1, characterized in that the data received in the frames are received in the form of digital radio signals. 4 »Method according to claim 1, characterized in that SOVA algorithm generates only flexible outputs (22) for positions without dotted lines in the convolution code. 5 »The method of claim 1, characterized in that in the flexible outputs (22) of the demodulator (17), the dotted positions are replaced by zeros. The method of claim 1, characterized in that the dotted line suppression (14) is only for outputting the rigid outputs (23). 7 »Method according to claim 1, characterized in that the values of the flexible outputs (22) also serve as criteria for stopping the iteration treatment. 8 "Device for decoding data transmitted in frames, the device having a demodulator (17) for the differential modulated data and a convolution decoder (20), characterized in that the demodulator (17) is a turbo demodulator (17), * at least one deinterleaver (12, 13) precedes the turbo demodulator (17), * the convolution decoder is an SOVA decoder (20). 9. The device as claimed in claim 8, characterized in that it comprises a memory made as an interleaving memory. A computer program product comprising program parts for executing the method according to at least one of the preceding claims 1 to 7 and a machine-readable data structure, in particular a computer, generated by a method according to at least one of claims 1 to 7 and / or with at least one computer program product as well as a machine readable data medium, in particular a computer readable data carrier having at least one registered computer program product. and / or stored and / or made available with at least one data structure. 13 11 °) Application of a device according to claims 8 and 9 to a radio receiver for decoding digital radio signals in data received in frames.