WO1999063700A1

WO1999063700A1 - Coded packet transmission without identifying the code used

Info

Publication number: WO1999063700A1
Application number: PCT/FR1999/001314
Authority: WO
Inventors: Frédéric GABIN; Stéphane GOSNE; Christophe Gruet; William Navarro; Philippe Thierion
Original assignee: Nortel Matra Cellular
Priority date: 1998-06-03
Filing date: 1999-06-03
Publication date: 1999-12-09
Also published as: CN1309854A; CA2334097A1; FR2779591B1; FR2779591A1; EP1084544A1; JP2002517946A

Abstract

The invention concerns a receiver for receiving a digital packet (B) having a transmission coding selected among a plurality of available coding systems, comprising decoding means (MD) for decoding said packet (B) according to the transmission coding. The transmission coding pertaining to a reduced set of possible coding systems, the receiver further comprises for each of the possible coding systems a decoder (DEC1, DEC2, DEC3) receiving one part of the packet to produce the dependability of the associated decoding, and it comprises means (COMP) for identifying the decoding means (MD) as those which correspond to the decoder having produced the best dependability. The invention also concerns a transmitter designed to co-operate with said receiver.

Description

The present invention relates to a method of transmitting digital packets which have been the subject of transmission coding, a method according to which the nature of the coding used is not transmitted.

The field of the invention is therefore that of digital transmissions by means of packets liable to have undergone different encodings, however all belonging to a set of available encodings. Thus, when a transmitter uses transmission coding to produce a packet from a message, it is important that the receiver for whom this packet is intended be able to identify this transmission coding to select the appropriate decoding means which will make it possible to recover the message. Although of very broad application, this field will be presented with reference to digital cellular radio systems of the GSM type. These systems have the advantage of being widely used and the support of a concrete example will make it possible to clarify the description of the invention.

According to current practice in digital telephony, an analog speech signal is digitized in 13-bit samples at the rate of 8 kHz, or a speed of 104 kilobits per second. GSM currently provides three types of source coding to reduce the bit rate of this digital signal. Full-speed coding, enhanced full-speed coding and half-speed coding, producing a signal at 13, 12.2 and 5.6 kilobits per second respectively from the previous signal.

Following the source coding which aims to compress the speech, the signal undergoes channel coding to protect it from the vagaries of radio transmission.

By considering the association of the source coding and the channel coding as a single coding, the transmission coding, the resulting signal has a bit rate of 22.8 kilobits per second in the case of full speed and 11.4 in the case of half speed.

This is the state of the art but it is already planned that future systems will use numerous transmission codings, these codings being able to be modified during communication depending on the quality of the radio link. The messages having a fixed length, it is generally provided, to minimize the technical complexity, that the different encodings produce packets of the same length. Thus, the sum of the bit rates of the source coding and the channel coding is constant. When the transmission channel is of good quality, it is possible to adopt a relatively low bit rate channel coding to favor the source coding, while otherwise, it is preferable to use a more robust channel coding to the detriment of the source coding. Naturally, the propagation conditions can change during communication, so that they may require a change in coding. It is therefore advisable to indicate to the receiver the nature of the coding which has been used for a given packet.

The immediate solution consists in reserving, within the packet, positions or mode bits to ensure this function. In this case, the receiver begins by detecting these mode bits to determine the decoding means suitable for the transmission coding which has been applied by the transmitter.

It goes without saying that these mode bits must also undergo special coding, mode coding, intended to ensure their protection during transmission. The mode coding, unlike transmission codings, must be unique so that the receiver can unambiguously identify the transmission coding used. The mode bits must therefore be coded independently of the useful content of the message which is subject to transmission coding. This mode coding is of course intended for the most difficult transmission conditions. severe and it is common to use a convolutional code in this case.

In terms of recall, such a code produces for a given bit, a number N of polynomials of degree K. Conventionally, we denote 1 / N the rate and K the constraint length of the code. By indexing a bit by its position in the message, a polynomial P associated with the bit bi, is defined by the coefficients aj, and takes the form of the following sum modulo 2: P = a ₀ bi + aibi-i + a ₂ bi_ ₂ + ... + a _k _ ₁ b _i _ _{k + 1} [2]

It is commonly accepted that to obtain satisfactory decoding, the minimum length of the coded word must be equal to five times the product of the constraint length by the inverse of the coding rate. It follows that for a rate 1/3 and for a constraint length equal to 5, appropriate typical values, the minimum size of the coded mode is 75 bits. We realize that if we provide 4 transmission codes, information which results in two bits for the mode, it is necessary to use 75 bits of the packet to transmit this information under the best conditions.

If the transmission efficiency is defined as the ratio of the number of bits carrying the information to be transmitted to the number of bits transmitted, it appears that this efficiency is far from optimal.

Thus, US Pat. No. 5,230,003 teaches a decoding system studied to distinguish signals encoded according to different available encodings, the encoding used not being transmitted. In this system, a decoder is required for each coding available. The number of decoders can become large when many codings are used.

The present invention thus relates to a method of transmitting coded packets which does not penalize the efficiency of transmission while limiting the complexity of the system. According to the invention, reception equipment is provided for receiving a digital packet which has been the subject of a transmission coding selected from among a plurality of available codings, and it includes decoding means for decoding this packet according to the coding of transmission; transmission coding belonging to a reduced set of possible codings, this equipment comprises for each of the possible codings a decoder receiving a part of the packet to produce the reliability of the associated decoding, and it further comprises means for identifying the decoding means as those which correspond to the decoder having produced the best reliability.

The invention also relates to transmission equipment intended to transmit a sequence of coded messages by means of packets, the last message of this sequence being subjected to a coding identified in a set of codings available and different from the coding applied to the first message of thereafter, these packets comprising on the one hand a section useful for receiving data and on the other hand guard bits, this equipment comprising means for arranging each of the coded messages in the whole of the useful section of the corresponding packet; moreover, the coding applied to the last message belongs to a reduced set of possible codings.

Preferably, the first packet of a transmission is assigned a predetermined available coding.

In addition, the possible encodings are the following available encoding, that which is identical to, and that which precedes the encoding of the preceding packet.

Advantageously, the possible codings are convolutional codings each affected by a separate coding scheme.

It is therefore desirable that the coding schemes be distinguished by the coding rate.

In addition, when the equipment is intended for reception, if the possible codings are among D

three, the identification of the decoding means can be carried out by means of two coding rates.

The invention will now appear in more detail in the context of the following description of embodiments given by way of example with reference to the appended figures which represent:

FIG. 1, the diagram of a receiver allowing the implementation of the invention, and

- Figure 2, the diagram of a transmitter for the implementation of the invention.

According to the invention, the mode which indicates the transmission coding to which a packet has been subjected is not transmitted by the transmitter.

In the exemplary embodiment which follows, four transmission codings are available which are each identified by a mode 1, 2, 3 and 4. Each transmission coding has an overall bit rate of 22.8 kilobits per second (kbpε) and associates source coding and channel coding; we give the following numerical example: - mode 1: source = 12.2 kbpε - channel = 10.6 kbps

- mode 2: source = 9.2 kbps - channel = 13.6 kbps

- mode 3: source = 7.8 kbps - channel = 15.0 kbps

- mode 4: source = 6.5 kbps - channel = 16.3 kbps

Mode selection is based on the estimated signal-to-noise ratio of the link between the transmitter and the receiver. This report therefore results from measurements made at the receiver and which are passed on to the transmitter so that the latter selects the appropriate transmission coding. The signal-to-noise ratio measurements are part of the state of the art so that they will not be more detailed.

Using the previous data, the transmitter selects one of the modes according to the signal to noise ratio C / I estimated as follows: - mode 1: C / I> 13 dB

- mode 2: 10 dB <C / I <13 dB

- mode 3: 7 dB <C / I <10 dB - mode 4: C / I <7 dB

Furthermore, following the source coding applied to a given source word, the convolutional channel codings of the different modes produce packets which have the following characteristics:

- mode 1: 318 bits at rate 1/2 followed by 138 bits at rate 2/3

- mode 2: 222 bits at rate 1/3 followed by 234 bits at rate 1/2

- mode 3: 384 bits in rate 1/3 followed by 72 bits in rate 1/2

- mode 4: 324 bits in 1/3 rate followed by 132 bits in 1/4 rate The receiver is designed to decode according to any one of the modes using the Viterbi algorithm. This algorithm produces, for an analyzed word, a decoded word as well as a metric. This metric indicates the distance between the analyzed word and a reference word which, subjected to this algorithm, produces the same decoded word. This metric is therefore a measure of the reliability of the decoding.

The maximum likelihood detection algorithm proceeds according to a fully specified coding scheme, namely in particular the rate of the code, the polynomials used and the position in the packet of the different coded bits. It calculates for different possible sequences of bits the metrics that they present with respect to the word analyzed in order to finally retain the sequence of bits affected by the highest metric. Thus, when this algorithm operates according to a coding scheme which does not correspond to the coding used for the analyzed word, the various sequences of bits will present metrics which are substantially similar. If, on the other hand, the chosen coding scheme is adapted to the word analyzed, a particular series of bits will have a much higher metric than the others, and this is therefore the solution sequence.

It will be specified that the deviation between the minimum metric and the maximum metric will be all the smaller as the coding parameters and those of the decoding will be strongly decorrelated. It is therefore necessary to select the channel codings of the different modes so that it has the lowest possible correlation. Several provisions can be adopted in this regard. First, a complete inversion of the bits of a packet can be provided, for example in modes 2 and 3.

Secondly, it is preferable to retain separate polynomials for each of the modes and to order them differently.

Third, it is advisable to adopt different coding rates, as far as possible.

The receiver will therefore take advantage of the disparities in the different channel codings to detect the transmission coding used in a received packet. To this end, it will try to decode this packet according to each of the channel codings to retain the one which has the highest metric at output.

It will first be noted that it is not necessary to decode the entire packet according to the four possible codes to obtain satisfactory detection. It suffices to proceed on a significant part of the package, the first part for example.

It will then be noted that the number of possible codes in a package can be limited compared to the four available codes. For example, a received packet can only be affected by the preceding mode, the same mode or the mode following that of the previous packet: a mode 4 packet may be followed by a mode 3 or 4 packet, and a mode 2 packet may be followed by a mode 1, 2 or 3 packet. It is further provided that the first packet received must be mode 4 so that there is no ambiguity at the start of the transmission.

Referring to Figure 1, the receiver will now be presented in more detail. This receiver includes a TRUNC truncation circuit which receives a packet B to keep part of it S, the first 138 bits in this case. The receiver keeps in memory the coding mode Pr of the previous packet.

It includes a first DECl decoder which decodes the part S of the packet according to the mode (Pr-1) to produce the corresponding metric Met (Pr-1).

It includes a second decoder DEC2 which decodes the part S of the packet according to the Pr mode to produce the corresponding metric Met (Pr).

It also includes a third decoder DEC3 which decodes this part S according to the mode (Pr + 1) to produce the associated metric Met (Pr + 1).

It will be noted here that when Pr is 1 the first DECl decoder is useless and we can in this case force

Set (Pr-l) to zero. Similarly, if Pr is 4 the third decoder DEC3 is of no interest and its output metric Met (Pr + 1) is also made zero.

On the other hand, one skilled in the art will note that the three decoders presented here as separate entities could very well be produced by means of a single processor provided for the processing of the Viterbi algorithm, this processor being parameterized according to the mode (Pr- 1), Pr or (Pr + 1) to ensure the respective functions of the first DECl, second DEC2 or third DEC3 decoder.

The receiver also includes a comparison circuit COMP which searches for the winning mode m having produced the highest metric:

Met (m) = Max [Met (Pr-l), Met (Pr), Met (Pr + l)]

As a precaution, it may be wise in the search for the winning mode m to ensure that it has produced a significantly stronger metric, twice for example, than the weakest metric. If this is not the case, it is reasonable to declare that the winning mode m is worth the previous mode Pr. In any event, if it is not possible to easily decide between the three decoders, it is very likely that the package concerned is unusable. This receiver naturally includes decoding means MD which receive the entire packet B to produce a word decoded by application of the Viterbi algorithm parameterized according to the winning mode m. Here again, these decoding means are not necessarily produced with an independent circuit. Advantageously, it is possible to reuse the processor possibly intended to replace the three decoders.

In addition, these decoding means may be limited to decoding the part of the packet which has not already been decoded by the decoder having produced the highest metric.

The general principle of the receiver having been revealed, we will now describe arrangements to this principle which take into account the specificity of the codes mentioned above.

It is easy to see that the three decoders can be replaced by two modules performing Viterbi decoding on 72 bits, the first according to a rate

1/3 producing a metric M3 and the second at a rate 1/2 producing a metric M2.

Similarly, the comparison circuit COMP can be simplified to now establish a differentiation value F signifying which of the two metrics M2, M3 wins. For example, by noting p a predetermined weighting coefficient, this differentiation value F takes the following values:

- if M3 - p.M2> 0, then F = 3

- if M3 - p.M2 <0, then F = 2

Thus, when the previous mode Pr is equal to 4, it suffices to analyze the first 72 bits of the packet with the two modules. If the differentiation value F is 3 the winning mode m is mode 4 while if this value is equal to 2, the winning mode is mode 3.

When the previous mode Pr is equal to 3, the two modules are loaded again with the first 72 bits of the packet. If the differentiation value F is 3, the winning mode m is the only possible one which has the rate 1/3, that is to say mode 4. If, on the other hand, the differentiation value is equal to 2, the two modules are now loaded with the following 72 bits of the packet. If the new differentiation value F is 3, the winning mode m is mode 3 while in the opposite case, it is mode 2 which is winning.

When the previous mode Pr is equal to 2, the 72 bits following the I38 ^th packet bit are taken into account. If the differentiation value F is 3, the winning mode m is the only possible one which has a rate 1/3, that is to say mode 3. If on the other hand the differentiation value is equal to 2, we load now the two modules with the 72 bits following the 234 ^th bit of the packet. If the new differentiation value is 3, the winning mode m is mode 2 and if not, it is mode 1 which is winning.

Finally, when the previous mode Pr is equal to 1, the two modules are loaded with the 72 bits which follow the 234 ^th packet bit. If the differentiation value F is 3, the winning mode is mode 2 while in the opposite case, it is mode 1 which is winning.

It thus appears that the invention can be implemented in many different ways which it is not possible to list exhaustively. The important point is to search on one or more parts of the packet for the mode which gives the best reliability in decoding, this for example by means of the corresponding metric.

The different modes are distinguished here by the coding rates which differ according to the position of the bit in the packet. One can also consider differentiating the modes by the coding polynomials assigned to them. We can also play on the position of the coded bits in the packet. In summary, the different modes should have a different coding scheme, whether it be the coding rate, the nature of the polynomials or the position of the coded bits. In addition, the invention applies regardless of the type of coding used and is not limited to convolutional codes. It is only important to be able to distinguish on reception, with good reliability, the nature of the coding of a received packet by searching for that of the possible codes from which it is most probably derived.

The invention further relates to a transmitter intended to transmit packets to the receiver.

This transmitter has the advantage of being simplified since it does not transmit the nature of the transmission coding used for the packet.

It should be recalled here that a packet results from the coding of the succession of a head section, a useful section, and a tail section. Indeed, the use of a convolutional code of constraint length K requires the use of (K-1) guard bits in the head section and the same number of guard bits in the tail section. The guard bits therefore frame the useful section. This useful section corresponds to the exploitable part, it being understood that the guard bits cannot be used to transmit information. Guard bits that are predetermined are used only during decoding. According to the invention, the entire useful section can be used to transmit the data which is the subject of the transmission between the transmitter and the receiver. The nature of the transmission coding is not included in the useful section, even when the coding has changed compared to the previous packet.

With reference to FIG. 2, the transmitter therefore comprises a control circuit CC which receives the nature N of the coding to be applied to the message W which should be conveyed by means of the next packet. It also includes a coding unit COD which receives this message W to code it as a function of the coding parameters Pa supplied by the control circuit CC. In this case, the control circuit CC produces the coding scheme based on the required channel coding.

The transmitter also includes a register ϋ which corresponds to the useful section of the packet. This register is loaded in full with the coded message MC originating from the coding member COD.

The other components of the transmitter will not be more detailed since they belong to the state of the art.

The implementation of the invention as set out above is of course only an example. Those skilled in the art have many possibilities for implementing the invention differently, if only by replacing one means with equivalent means.

Claims

1) Reception equipment designed to receive a digital packet (B) having been the subject of a transmission coding selected from a plurality of available codings, comprising decoding means (MD) for decoding said packet (B) according to said transmission coding, characterized in that, said transmission coding belonging to a reduced set of possible codings, it comprises for each of said possible codings a decoder (DECl, DEC2, DEC3) receiving a part of said packet to produce the reliability of the associated decoding , and it further comprises means (COMP) for identifying said decoding means (MD) as those which correspond to the decoder having produced the best reliability.

2) Transmission equipment provided for transmitting a sequence of coded messages by means of packets, the last message (W) of this sequence being subjected to a coding identified in a set of available codings and different from the coding applied to the first message of the suite, these packets comprising on the one hand a useful section (U) for receiving data and on the other hand guard bits, this equipment comprising means (CC) for arranging each of said coded messages (MC) in the whole of the useful section (U) of the corresponding packet, characterized in that the coding applied to said last message belongs to a reduced set of possible codings.

3) Equipment according to any one of claims 1 or 2, characterized in that the first packet of a transmission is assigned a predetermined available coding.

4) Equipment according to claim 3, characterized in that said possible encodings are the available encoding which follows (Pr + 1), that which is identical to (Pr), and that which precedes (Pr - 1) the encoding of the preceding packet (Pr). 5) Equipment according to any one of the preceding claims, characterized in that said possible codings are convolutional codings each assigned a separate coding scheme. 6) Equipment according to claim 5, characterized in that said coding schemes are distinguished by the coding rate.

7) Equipment according to claim 6 characterized in that, intended for reception, said possible codings being three in number, the identification of said decoding means (MD) is carried out by means of two coding rates.