DE10219151A1 - Adapting data rate in communications device involves removing (dotting) certain transmission using dotting pattern configured to remove preferred transmission bits depending on coding process - Google Patents

Abstract

The method involves dividing the data stream into at least one data block containing transmission bits, forming the transmission bits from input bits carrying information in a coding process and removing (dotting) certain transmission bits to adapt the data rate using a dotting pattern configured to remove preferred transmission bits depending on the coding process for at least some input bits. AN Independent claim is also included for the following: (a) a communications device with a rate adapter.

Description

The present invention relates to a method according to the Preamble of claim 1 and 11 for adjusting the Data rate in a communication device and a corresponding communication device according to the preamble of claim 16.

Different applications in communication systems mostly work with different data rates. The underlying transmission channels usually offer, for. B. because of the embedding in certain transmission formats, only a fixed data transfer rate or raw data transfer rate or only a discrete set of such data rates. It will therefore generally be necessary to adapt the data rates to one another at the corresponding interface. This is described in the following using an example from UMTS standardization:
Work is currently underway to standardize the so-called UMTS mobile radio standard ("Universal Mobile Telecommunication System") for mobile radio devices of the third mobile radio generation. According to the current state of UMTS standardization, it is provided to subject the data to be transmitted via a high-frequency channel to channel coding, with convolutional codes being used in particular for this purpose. The channel coding encodes the data to be transmitted redundantly, which enables a more reliable recovery of the transmitted data on the receiver side. The code used in the channel coding is characterized by its code rate r = k / n, where k denotes the number of data or message bits to be transmitted and n the number of bits present after coding. The lower the code rate, the more efficient the code is. However, a problem associated with coding is that the data rate is reduced by the factor r.

To the data rate of the encoded data stream to the to adjust the possible transmission rate, a Rate matching ("rate matching") carried out, after a certain pattern either bits from the data stream removed or doubled in the data stream. The Removing bits is called "puncturing" and doubling referred to as "repeating".

According to the current state of UMTS standardization proposed an algorithm for rate adjustment use a dotting with an approximation performs regular puncturing patterns, d. H. the too puncturing bits are equidistant from each other punctured, coded data block distributed.

It is also known that the convolutional coding Bit error rate (BER) at the edge of a correspondingly coded data blocks decreases. Likewise known that the bit error rate within a data block locally changed by unevenly distributed puncturing can be. It is also from WO 01/26273 A1 and WO 01/39421 A1 known that it is beneficial to the individual Data blocks of the data stream for adapting the data rate according to to puncture a certain dotting pattern, the Dotting pattern is designed such that it is one of a middle range of the individual data blocks at least one end of the individual data blocks steadily has increasing puncturing rate.

The present invention is therefore based on the object a method for adjusting the data rate of a data stream in a communication device as well as a corresponding one To provide communication device which a ensure satisfactory bit error rate and especially in mobile radio systems with convolutional coding can be used.

This object is achieved by a method with the features of claims 1 and 16 or by Communication device with the features of claim 16 solved. The sub-claims define preferred and advantageous embodiments of the present invention.

The systematic of the convolutional code was used to to find heuristic puncturing patterns according to their Application all bits of the punctured data block one of their have the corresponding bit error rate for each importance.

Preferably, the puncturing pattern has one of those middle area at both ends of the respective data block increasing puncturing rate. That way the bits at the beginning and end of each to be punctured Data blocks more punctured, but not with a uniform puncturing rate, but with an im Essentially to the two ends of the respective data block the puncture rate increases, d. H. the distance between the dotted bits becomes the two ends of the Data blocks are getting shorter and shorter on average. As below the puncturing rate must be executed Surprisingly, not necessarily strictly monotonous Rising ends, or in other words the Reduce puncturing distance strictly monotonously. Rather, it can it, due to the specific properties of the convolutional codes used and in particular those used Generator polynomials, too, may be beneficial to something to use more irregular patterns.

This puncturing leads to one above the punctured one Data block more evenly distributed error rate of each Bits and also has a diminished Overall probability of error.

The present invention is particularly suitable for Adaptation of the data rate of a convolutionally coded data stream and can therefore preferably be used in UMTS mobile radio systems be, this being both the area of the mobile radio transmitter as well as that of the mobile radio receiver. The invention is however not limited to this area of application, but can generally be used wherever the Data rate of a data stream is to be adjusted.

The present invention is hereinafter referred to on the attached drawing based on preferred Exemplary embodiments described in more detail. Show it:

Fig. 1 is a simplified block diagram of a mobile radio transmitter according to the invention.

Fig. 2 shows the bit error rate BER per bit for puncturing according to an embodiment of HS-SCCH, Part 2, coding with R = 1/3 with a ratio of the energy of the transmitted bits to the noise power density E _s / N ₀ = -2 dB. The HS-SCCH channel is the so-called high-speed shared control channel, via which certain configuration information is transmitted and which can be divided into two parts, the so-called Part 1 and Part 2. Part 1 is transmitted first and contains the information that the mobile station needs first to process the following data channel; Part 2 contains information that the mobile station only needs a little later. This division in two means that the delay through the HS-SCCH is as small as possible, since only the first part has to be decoded before the data can be received.

Fig. 3, the BER per bit for the in UMTS (Specification 25.212 v5.0.0, chap. 4.2.7 "rate matching") rate matching proposed in the HS-SCCH, Part 2, at a ratio of the energy of the transmitted bits to the noise power density of E _s / N ₀ = -2 dB.

Fig. 4 shows a comparison of with an inventive puncturing (upper curve, crosses) or a conventional puncturing (lower curve, circles) with respect to the resulting total error probability achievable results, wherein at least one bit of a block has been transmitted erroneously, the probability here ( so-called frame error rate) is plotted.

Fig. 5 underlying schemes for convolutional codes in UMTS.

Fig. 6, the BER per bit for the in UMTS (Specification 25.212 v5.0.0, chap. 4.2.7 "rate matching") proposed rate matching in the HS-SCCH, Part 1, wherein a ratio of the energy of the transmitted bits to the noise power density of E _s / N ₀ = -3 dB.

Fig. How many input bits (input bits) are affected 7 at a dotting of output bits in the different output stages Output 1, Output 2 and Output 3.

Fig. 8 which input bits (bit numbers) are affected by the puncturing.

Fig. 9 is a table showing the results of the puncturing function of the number of punctured bits.

Fig. 10, the bit error rate BER per bit for puncturing according to an embodiment in the HS-SCCH, Part 1, at a signal-to-noise ratio of the energy of the transmitted bits to the noise power density of E _s / N ₀ = -3 dB.

Fig. 11, various embodiments for puncturing of 8 bits (48 to 40 bits) for a coding rate 1/3.

Fig. 12 different embodiments for puncturing 31 bits (puncturing 111 to 80 bits), R = 1/3.

Fig. 13 different embodiments for a repetition of 31 bits (repetition from 32 to 40 bits), R = 1/2.

Fig. 14 different embodiments for a repetition of 6 bits (74 to 80 bits), R = 1/3.

Fig. 15 different embodiments for a repetition of 4 bits (36 to 40 bits), R = 1/3.

Fig. 16 different embodiments for puncturing 14 bits (54 to 40 bits), R = 1/3.

Fig. 17 show further embodiments for a puncturing of 31 bits (dotting from 111 to 80 bits), R = 1/3. This figure can therefore also be regarded as a continuation of FIG. 12.

Fig. 18 shows an embodiment of a puncturing from 108 to 80 bits, R = 1/3.

Fig. 19 embodiments for puncturing 114 to 80 bits, R = 1/3.

Fig. 20 embodiments for puncturing 117 to 80 bits, R = 1/3.

Fig. 21 embodiments for puncturing from 52 to 40 bits, R = 1/2.

Fig. 22 embodiments for puncturing from 46 to 40 bits, R = 1/2.

Fig. 23 embodiments for puncturing from 54 to 40 bits, R = 1/3.

Fig. 24 embodiments for puncturing from 56 to 40 bits, R = 1/2.

Fig. 25 embodiments for a repetition of 36 to 40 bits, R = 1/2.

Fig. 26 embodiments for puncturing from 48 to 40 bits.

Fig. 27 embodiments for puncturing from 11 to 80 bits.

Fig. 28 Rate adjustment regulation from the 3-GPP specification 25.212 v5.0.0, chap. 4.2.7 "Rate matching".

In general, the lines in the tables with numbers printed entirely in bold mean the respectively particularly preferred exemplary embodiment, but the quality of the other exemplary embodiment does not necessarily deviate significantly from this emphasized exemplary embodiment. In FIGS. 26 and 27, however, numbers in bold denote the bits punctured or repeated by the described construction principle of the rate adjustment formula according to the invention at the beginning or end of the repetition pattern. These are thus fixed, whereas the position of the bits not shown in bold can also be shifted slightly (typically by one position) by varying the parameters in the context of this invention.

In Fig. 1 shows the structure of a mobile station 1 according to the invention is schematically shown, are transmitted from the data or communication information, particularly voice information via a radio frequency transmission channel to a receiver. In Fig. 1, in particular the components involved in the encoding of this information or data are shown. The information supplied by a data source 2 , for example a microphone, is first converted into a bit sequence using a digital source encoder 3 . The speech-coded data are then encoded with the aid of a channel encoder 4 , the actual useful or message bits being encoded redundantly, as a result of which transmission errors can be identified and then corrected. The channel encoder 4 can be a convolutional encoder. The code rate r resulting in the channel coding is an important variable for describing the code used in each case in the channel coding and, as has already been mentioned, is defined by the expression r = k / n. Here, k denotes the number of data bits and n the number of bits coded in total, ie the number of redundant bits added corresponds to the expression n-k. A code with the code rate r defined above is also referred to as an (n, k) code, the performance of the code increasing with decreasing code rate r. So-called block codes or convolutional codes are usually used for channel coding.

In the following, it should be assumed that - as determined by the current state of UMTS standardization - convolutional codes are used for channel coding. An essential difference from block codes is that convolutional codes do not encode individual data blocks one after the other, but that they are continuous processing, with each current code word of an input sequence to be coded also depending on the previous input sequences. Regardless of the code rate r = k / n, convolutional codes are also characterized by the so-called influence length or "constraint length" K. The "constraint length" specifies the number of clock cycles of k new input bits of the channel encoder 4, a bit influences the code word output by the channel encoder 5 .

The following convolutional codes are used for UMTS, as shown in FIG . The illustration is from specification 25.212, chap. 4.2.3.1 "Convolutional coding" (convolutional codes) taken.

Before the channel-coded information is transmitted to the receiver, it can be fed to an interleaver 5 , which rearranges the bits to be transmitted according to a certain scheme and spreads them over time, thereby distributing the errors that generally occur in bundles in order to create a so-called memoryless ) Receive transmission channel with a quasi-random error distribution. The information or data coded in this way is fed to a modulator 7 , the function of which is to modulate the data onto a carrier signal and to transmit it to a receiver via a high-frequency transmission channel 3 in accordance with a predetermined multiple access method.

For transmission, the coded data stream is divided into data blocks, the channel encoder 4 being set to a known state at the start of a data block. In the end, each coded data block is terminated by so-called “tail bits”, so that the channel encoder 4 is again in a known state. This construction of the convolutional code and of the channel encoder 4 ensures that the bits at the beginning and end of an encoded data block are better protected against transmission errors than in the middle of the block. It is irrelevant whether these tail bits all have the known value 0 or another value. The values of these tail bits can also be chosen arbitrarily, whereby both sender and receiver must know the values to be used.

The error probability of a bit depends on its location within the respective data block differently. This effect is used for example in the Voice transmission in GSM mobile radio systems exploited by the most important bits are placed at the two ends of the block where the probability of error is lowest. When data transfers, however, in general Data packets are discarded if only one transmitted bits is faulty, for example in Receiver through a so-called "Cyclic Redundancy Check" (CRC) can be determined. Therefore, with a Data transfer from important or less important Bits are spoken, but all bits are considered the same important to look at. If errors in a control block occur, i.e. a data block, the control information contains what contains information about how subsequent user data are encoded and transmitted generally correct detection of this useful data then no longer possible if only a single bit is received incorrectly because the received data is then incorrect be interpreted.

In order to adapt the data rate of the coded data stream to the respective possible transmission rate, a rate adjustment ("rate matching") is carried out in front of the modulator 7 . In the exemplary embodiment shown in FIG. 1, the rate adjustment takes place in the rate adjustment unit 6 b, the puncturing unit 6 a first performing puncturing in accordance with a specific puncturing pattern in order to achieve a more uniform error distribution over a data block. The sequence of the puncturing unit 6 a and the interleaver 5 shown in FIG. 1 are only to be understood as examples. The interleaver can also be arranged after the unit 6 b. Also, the interleaver 5 can also be by two interleaver before and after the rate matching unit 6 b replaced etc.

It is also an object of this invention To further optimize puncturing patterns and in particular to adjust the polynomials used for the channel encoder. It So the task arises, depending on the one used Convolutional code (including the polynomials used) and the block length the amount of the to be punctured or select repetitive bits so that the decoding can be carried out as well as possible. Usually result yourself a large number of ways, so at least there is is very time and resource consuming, a very good one Develop rate adjustment patterns purely through simulation. For example, if you want all possible puncturing patterns to puncture from 48 bits to 40 bits, so would be the 48! / (8! .40!) = 377348994 different ways that cannot be examined in a reasonable time.

This problem arises particularly for short block lengths, such as. B. for the control information of the UMTS extension HSDPA, and there in particular the HS-SCCH channel (High Speed Shared Controll CHannel). This channel transmits configuration information which specifies how the actual user data sent via a special data channel is coded and further details, e.g. B. the spreading codes used for transmission. In contrast to the data channel, over which a great deal of data can be transmitted, this is comparatively little data. In UMTS, convolutional codes with the rate 1/2 or 1/3 are used for coding, the polynomials used are shown in FIG. 5. Polynomials are also the exact design of the "tapping points", that is to say which delay stages for the individual output bit streams are tapped and linked by an exclusive-OR operation.

The invention is therefore particularly applicable to the so-called HS-SCCH (High Speed Shared Controll CHannel).

The definition of the coding of the HS-SCCH is based on the current State of the art in the specification 3GPP TS 25.212 V5.0.0 (2002-03), "Multiplexing and channel coding (FDD) (Release 5) ", especially in chapter 4.6" Coding for HS- SCCH ". This specification is otherwise mentioned in this application also referred to as 25.212. In subchapter 4.6.6 "Rate matching for HS-SCCH "states that the Rate adjustment according to the standard rate adjustment algorithm in Chapter 4.2.7 "Rate matching" must be carried out, which is essentially equidistant (if possible) puncturing or repetition.

• a) 32 auf 40 (mit Code-Rate R = 1/2) oder
• b) 48 auf 40 (mit Code-Rate R = 1/3) und
• c) 74 auf 80 (mit Code-Rate R = 1/2) oder
• d) 111 auf 80 (mit Code-Rate R = 1/3).

The block length of the two parts of the HS-SCCH is currently 8 bits for the first part, or if the end bits (tail bits) are included 16 bits, 29 bits for the second part, or if the end bits (tail bits) are also included become 37 bits. Since the specification is still in flow, changes in various parameters can also result in other block lengths. Furthermore, the convolutional codes with the rate or 1/3 can also be used. The following rate adjustments are particularly relevant:

• a) 32 on 40 (with code rate R = 1/2) or
• b) 48 on 40 (with code rate R = 1/3) and
• c) 74 to 80 (with code rate R = 1/2) or
• d) 111 to 80 (with code rate R = 1/3).

Procedure for determining puncturing and Repetierungsmustern

In the overview, it can thus be found that puncturing and / or repetition or even repetition is carried out in such a way that the overall bit error rate (BER) is minimal. The situation in FIG. 3 is first considered here: the bit error rate for the individual bits is plotted in a frame. The abscissa represents the index of the respective bit "(frame index)". It can clearly be seen that the first and last bits have a lower bit error rate. This can be understood in connection with the scheme for convolutional codes from FIG. 5: For the transmission, bits from the different delay stages D of the decoder are linked together by the convolutional code. The first bits are e.g. B. also linked with the bits preceding them, that is to say bits that actually do not exist. These "non-existent bits" are then set to a known value, usually zero. This is known to the receiver, who in turn decodes the first transmitted bits with these bits set to zero. Decoding is very safe here, since some of the bits are known with absolute certainty.

The same applies to the last bits: following them become artificial bits again, the so-called end bits or "tail bits", in the delay elements D of the decoder be inserted; these end bits are in turn one known value, usually set to zero.

In the middle area bits are linked with each other Value at the recipient is not known with certainty. So is when decoding, the greater the probability that a Error occurs, which results in a higher bit error rate manifests.

• a) die Hüllkurve stellt im Wesentlichen eine Horizontale dar (oder nähert sich ihr an):
  Das bedeutet, dass die Bitfehlerrate für alle Bits innerhalb eines Rahmens im Wesentlichen gleich ist. Dies geschieht beispielsweise durch eine Punktierung am Rande oder eine Repetierung in der Mitte, oder beides, auch abhängig davon, auf welche Rate angepasst werden soll.
• b) konkavartige Ausbildung der Hüllkurve:
  In diesem Falle wird beispielsweise am Rand so stark punktiert, dass die Bits im mittleren Bereich des Rahmens eine geringere Bitfehlerrate aufweisen. Dieser Sachverhalt ist in Fig. 2 zu sehen.
• c) die Bitfehlerrate ist unregelmäßig gegenüber der Rahmennummer verteilt. Dieser Fall wird aus weiter unten ausgeführten Gründen hier nicht näher betrachtet.

The envelope of the bit error rate in relation to the frame number is here initially shaped convexly with even repetition or puncturing. There are now different ways in which the envelope changes when the puncturing (or repetition) is changed:

• a) the envelope curve essentially represents (or approximates) a horizontal:
  This means that the bit error rate is essentially the same for all bits within a frame. This is done, for example, by puncturing on the edge or repeating in the middle, or both, depending on the rate to which adjustment is to be made.
• b) concave design of the envelope:
  In this case, the edge is punctured so strongly, for example, that the bits in the central area of the frame have a lower bit error rate. This fact can be seen in Fig. 2.
• c) the bit error rate is distributed irregularly with respect to the frame number. This case is not considered here for reasons explained below.

The following explanations relate to puncturing. Analogous considerations can also be made for repeating or for combinations of repeating and puncturing:
There are now many ways in which individual bits can be punctured. If, for example, as already explained above, one would like to examine all possible puncturing patterns for puncturing from 48 bits to 40 bits, this would be 48! / (8! .40!) = 377348994 different possibilities, which cannot all be examined in a reasonable time ,

The aim is therefore not sensible from the outset Ways to eliminate. This doesn't happen over any repetition and / or puncturing, why Alternative c) is not considered further here.

An ordering principle is shown in FIG. 7. For the first 9 input bits 1-9 and the last 9 input bits n-8 to n, the puncturing level for the respective output stage Output 0, Output 1, Output 2 is shown. The output stages themselves, as can be seen in FIG. 5, are the respective output functions which are formed from all input bits preceding the input bit under consideration by linking. The output stages of FIG. 5b) are considered here, that is to say the rate 1/3 convolutional encoder. For puncturing with as little information loss as possible, it is advisable to omit bits (puncturing) that have little influence on other bits. The puncturing level therefore indicates how many bits are influenced by puncturing the bit under consideration.

An exemplary procedure for omitting or puncturing bits is shown in FIG. 8. The first 9 input bits 1-9 and the last 9 input bits n-8 to n are shown in the first column. In the following columns, the bit numbers of the information bits affected by the puncturing, i.e. information bits or input bits, are output for the respective output stage 0, Output 1 and Output 2 shown. As in FIG. 7, the table fields have an increasingly dark background as the number of information bits affected increases. The bits belonging to the bright table fields are therefore candidates for puncturing.

A table is shown in FIG. 9, in which the important variables when puncturing near the ends, ie puncturing the first and last bits, are illuminated. N input bits (information bits) and k coded bits (bits at the output stage, output bits) are considered. In the first column the number of punctured output bits (# punct bits) is given, in the last column (cumulative) the number of information bits affected at the input, whereby input bits that are affected several times, i.e. by puncturing several output bits, also be counted accordingly several times.

In the second column, under Sequence, it is specified which Output bit (bit number) was punctured in this step. The puncturing is done starting with the least important bits in the first line, towards the following bits in the following lines. The overall dotting pattern for z. B. 7 bits to be punctured results from the in Column 2 in the lines 1 to 7 specified bits, i.e. the bits 1, k, 4, k-4, k-6, 2, k-1. So this pattern includes the Bits 1, 2, 4, k-6, k-4, k-1, k.

The indexing for is above the first line the first information bits 1-9 and the last Information bits k-8 to k. For space reasons, instead of k-8 only written -8 etc. The entries in the columns below the indexing of the information bits indicate how strong that is relevant information bit by puncturing the Output bits in the 2nd column up to the respective line are specified and are therefore punctured, is affected. The means how many of the punctured output bits were with linked to this information bit. That's a measure of how strong the information bit in question by the puncturing was weakened.

Finally, in the last column (cumulative) is the sum of these impairments listed. It will be here called cumulative puncture strength.

The column mean gives the ratio V to the sum of the last column divided by the number of affected Information bits. For example, for 6 punctured bits V = (2 + 1 + 1 + 1 + 1) / (1 + 1 + 1 + 1 + 1) = 1.2 The average puncturing rate (average P rate) is the column "Average" divided by 18, the total number of each Information bit of "exclusive" or "linkages.

A way to get any number of bits dot, consists of tables analogous to those mentioned above to customize. For rate 1/3 and the polynomials considered Convolutional encoder can use the tables shown become. At other coding rates and / or others Polynomials can be easily determined analogously in the tables. With the help of these tables you can then create one Puncturing order determines where such output bits first be punctured that have little impact on the have cumulative puncturing strength. There are several Alternatively, such bits are preferably punctured, which is the maximum of the puncturing strength of the individual bits minimize.

For a higher number of bits to be punctured and / or larger Block lengths must i. d. Usually the information from the tables with the idea of being distributed as evenly as possible across the whole Dotting block can be combined. Then it offers itself to puncture additional bits in the middle part, those from the generator polynomial with the least powers, d. H. With few links are generated. simultaneously However, it must be ensured that the overall distribution of the Dotting strength in the middle area of the frame none has significant increases.

The same applies to repetition, in each case with the reverse Sign. That is, bits that according to the heuristic would be punctured first, now repeated and that Generally, an even repetition in the middle section first is carried out, preferably by the polynomials with the most shortcuts. Then such bits are on the edge repeats that (when punctured) the largest possible Influence the cumulative puncturing strength.

Unlike procedures where the puncturing rate increases steadily towards the ends, this leads to one an unexpected result because you would expect that the reliability of the coded bits towards the ends steadily increasing. But it shows up on closer inspection of the polynomials for the convolutional encoder used that this Surprisingly, the assumption is not correct. Through the specific properties of the polynomials arise, especially at the end, encoded bits that are less effective at Contribute coding. However, these bits do not come to an end increasingly increasing, but are something irregularly distributed. By using the dotting pattern specifically aimed at these "weak" bits preferably punctured these bits, the coding can still be used continue to improve.

• - mittels einer neu definierten, heuristischen Metrik die Auswirkung der Punktierung/Repetierung eines codierten Bits auf die zugrunde liegenden Informationsbits näherungsweise zu ermitteln,
• - gezielt und für jeden Faltungscode spezifisch Bits auszuwählen, die punktiert bzw. repetiert werden sollen,
• - die Anzahl der zu untersuchenden Ratenanpassungsmuster stark einzuschränken.

The invention therefore uses a heuristic method which allows:

• - to approximately determine the effect of puncturing / repeating a coded bit on the underlying information bits using a newly defined, heuristic metric,
• selectively select bits for each convolutional code that are to be punctured or repeated,
• - severely restrict the number of rate adjustment patterns to be examined.

After, based on this procedure, a few promising rate adjustment patterns have been identified are based on the frame error rate and the Bit error rate of each individual information bit (hereinafter referred to as the bit error rate distribution). iterative Then, based on the developed metric, the Rate adjustment patterns can be further refined and optimized. The bit error rate distribution of the non-punctured / not repeated blocks

The puncturing strength S _i per bit of information bit i is defined as a heuristic metric as the number of links of an information bit not transmitted by the puncturing to the respective output bits of the encoder. S _i is therefore positive for puncturing. For repetition, S _{i, k} = n-1 is defined for each n-fold link.

S _max is the maximum possible puncturing strength, given by the code-specific total number of existing links:

• 1. Wähle die kumulative Punktierungsstärke nahe beim möglichen Minimum.
• 2. Sorge für eine möglichst gleichverteilte Bitfehlerrate über allen Informationsbits.

A good rate adjustment pattern is sought according to the following quality criteria:

• 1. Choose the cumulative punctuation strength close to the possible minimum.
• 2. Make sure that the bit error rate is distributed as evenly as possible over all information bits.

For the selection of the bits to be punctured / repeated are based on the generator polynomials of the code for the beginning and end of the coded block tables established, which is the cumulative puncturing strength per coded bit, as well as the affected information bits represent. This allows the coded bits to be Classify the cumulative puncture strength classes.

According to the quality criterion above, we will now use this Tables to be punctured / repeated selected that initially for those information bits, that show a lower bit error rate than other bits that Puncturing strength is increased and at the same time the cumulative punctuation strength is kept low. The Puncturing strength is therefore inversely proportional to Bit error rate of the information bit is selected, and also specifically selected bits that have little to do with the cumulative Contribute puncturing strength.

This procedure is then based on the first determined pattern applied iteratively, so that already after a few simulations for the respective convolutional code specifically optimized rate adjustment patterns can be found can.

In Figs. 11 and 12 different possibilities are shown for puncturing according to the invention, wherein in each case the numbers of the (start counting at 1) to be punctured bits are indicated. The tables are given for different numbers of information bits to be transmitted and different numbers of bits to be transmitted after the rate adjustment.

In Fig. 3 the variation of the bit error rate for each transmitted bit of a data block, depending on their position or location in the data block for a conventional puncturing is exemplarily applied to a regular puncturing pattern.

In FIG. 2, this curve is shown for an inventive puncturing with the pattern number 33 of FIG. 12 which has been found to be particularly useful in simulations. It can be seen from FIG. 2 that by using the puncturing pattern according to the invention, a more uniform course of the bit error rate can be achieved over the entire data block. Since puncturing is less frequent in the central area of the data block compared to the conventional procedure, a lower probability of error can be obtained there. In fact, the error rate now increases somewhat towards the ends, which at first glance might seem unfavorable. However, this results from the fact that there are a particularly large number of “weak” bits at the edge, as already explained above, where puncturing can be carried out very cheaply.

In FIG. 4, the course of the total error rate is plotted against the ratio of the energy of the transmitted bits to the noise power density for the same cases. It can be seen from FIG. 4 that with the aid of the invention (lower curve, circles), a frame error rate which is improved by approximately 0.2 dB compared to the conventional procedure (upper curve, crosses) can be achieved.

Similar improvements can also be achieved with other parameters. For example, FIG. 6 shows the course of the bit error rate for the individual transmitted bits of a data block as a function of their position in the data block for conventional puncturing with a regular puncturing pattern with coding with rate 1/3 and puncturing 8 bits (48) 40 bits). This corresponds to a transmission of 8 input bits. FIG. 10 shows the distribution if, instead, the puncturing pattern No. 3 from FIG. 11 is used, which has also been found to be particularly suitable in simulations. You can see that there is a very balanced distribution here. Here, too, there is an improvement of approx. 0.2 dB (however, no curve has been added for this, since it does not bring any further significant findings).

Fig. 16 shows further preferred embodiments within the scope of the invention with puncturing of 14 bits of 54, wherein the patterns 3 and 4 get the best results.

Figures 13, 14 and 15 show preferred repetition patterns which were also obtained using the rules shown in this invention.

The present invention has been described with reference to use in a mobile radio transmitter. Of course, the invention can, however, also be extended to mobile radio receivers, where a signal punctured or repeated in order to adapt the data rate in the manner described above has to be processed in accordance with the puncturing or repetition pattern used in each case. In this case, additional bits are inserted into the received bit stream in the respective receiver for punctured or repeated bits on the transmission side, or two or more bits of the received bit stream are combined. When additional bits are inserted, it is noted in the form of a so-called "soft decision" information that their information content is very uncertain. The processing of the received signal can take place in the respective receiver in the reverse order to FIG. 1.

Others using the procedure outlined above determined rate adjustment patterns

The previously specified puncturing patterns concentrate predominantly on a puncturing in the end areas or / and a repetition in the middle range.

The other rate adjustment patterns now described have been with that in the previously explained procedure for various Proposals for HS-SCCH coding in standardization determined. The points to be punctured or bits to be repeated. The bits are from 1 to N numbered. The preferred pattern comes first called; the other patterns are always similar favorable properties.

FIG. 17, in which these further puncturing patterns are listed, therefore represents a supplement to FIG. 12. Correspondingly, puncturing patterns for different output bit rates are shown in FIGS. 18-24 and further repetition patterns in FIG. 25.

Approximation of preferred rate adjustment patterns below Use of components already specified in UMTS

The patterns shown so far have the goal of one if possible optimal selection of to be punctured or repeated Propose bits, but no other Pattern restrictions are assumed. In practical implementations it can be advantageous to consider only those patterns that are possible with small changes to existing rate adjustment circuits have implemented. A corresponding Rate adjustment regulation is in the already mentioned Specification 25.212 v5.0.0, chap. 4.2.7 "Rate matching" (Rate adjustment). The following is the part this rule, which is the actual puncturing or Repeats and in Chapter 4.2.7.5 "Rate matching pattern determination " Rate adjustment) is described, analogously reproduced.

Extract from the specification

Before the rate adjustment, the bits are marked with x _i1 , x _i2 , x _i3 , ..., x _{iX i} . Here i represents the transport channel number, the sequence itself is in sections 4.2.7.4 of the specification for the uplink and in 4.2.7.1. defined for the downlink. The uplink is understood to be the connection of a communication device to the base station, the downlink is the connection of a base station to a communication device.

• - Es wird zunächst eine Fehlerwert e auf einen Anfangswert gesetzt, welcher zwischen dem ursprünglichen Fehlerwert und der gewünschten Punktierungsrate liegt.
• - In einer Schleife mit dem Index m des momentan betrachteten Bits als Laufparameter wird bis zum Ende der Sequenz, also bis zum Index X_i.
• - Zunächst wird der Fehlerwert e auf e - e_minus gesetzt, wobei e_minus im Wesentlichen die Anzahl der zu punktierenden Bits darstellt.
• - Anschliessend wird überprüft, ob der Fehlerwert e <= 0 ist.
• - In diesem Fall wird überprüft, ob das Bit mit dem Index m punktiert werden soll, wobei ein zu punktierendes Bit dann auf einen Wert δ gesetzt wird, der verschieden von 0 oder 1 ist.

The rule for rate adjustment is in the figure ??? reproduced program section shown, which expires when the condition is met that puncturing is carried out:

• - First, an error value e is set to an initial value that lies between the original error value and the desired puncturing rate.
• - In a loop with the index m of the currently considered bit as a run parameter, until the end of the sequence, that is to say the index X _i .
• - First, the error value e is set to e - e _minus , where e _minus essentially represents the number of bits to be punctured.
• - It is then checked whether the error value e <= 0.
• In this case, it is checked whether the bit should be punctured with the index m, a bit to be punctured then being set to a value δ that is different from 0 or 1.

In the event that a repetition should take place, a essentially analogous process taking place, then a repeated bit set directly after the original bit becomes.

bit

With a puncture then in the further course removed the bits that were set to 8 so that so these bits are punctured.

The parameters X _i , e _ini , e _plus and e _minus are each selected so that the desired rate adjustment can be achieved. The following then essentially applies: e _plus = X _i , e _minus = N _p , where X _i denotes the number of bits before the rate adjustment and N _p the number of bits to be punctured or repeated. In principle, e _ini can be chosen anywhere between 1 and e _plus , with a slight shift in the pattern; this is used in certain cases (rate adjustment after a first interleaving) to suitably shift the patterns against each other in different frames. The parameter i identifies different transport channels in the specification. However, this parameter is irrelevant in the present case and is therefore omitted. In the following, possibilities are shown how one can approximate preferred rate adaptation patterns for short block sizes for convolutional codes by means of this already existing rate matching algorithm. Under the boundary condition of this algorithm, attempts are made to use bits at the ends of the code block when puncturing and, in particular, bits from the center of the code block when repeating. A key aspect of this exemplary embodiment is not to restrict the parameter e _ini to the value range from 1 to e _plus , but instead to select it advantageously outside this range. At first glance, such a choice may seem contradictory because it is no longer ensured that the desired number of bits will be punctured or repeated. However, by advantageously adapting the values of e _plus and e _minus , the desired number can still be achieved.

It is:
X _i : Number of bits before rate matching.
N _p : Number of bits to be punctured / repeated (the index _p in N _p indicates the number of bits to be punctured; N _p can also denote the number of bits to be repeated).

In order to fully specify the application of the rate matching algorithm and thus the rate adjustment pattern, the error start value e _ini , the error increment e _plus and the error decrement e _minus must be specified, since these parameters completely describe the rate adjustment pattern.

In the following the approximation of preferred Rate adjustment patterns using the UMTS rate in Release 99 Matching algorithm shown.

The following shows the options for using the Rate matching algorithm already in the standard (Data rate adjustment algorithm) preferred Rate adjustment patterns for short block sizes in convolutional codes can be approximated. It is under the Boundary condition this algorithm tries when puncturing prefers to use bits at the ends of the code block and with repetition, especially bits from the middle of the code block.

dotting

The parameters of the rate matching algorithm are chosen so that the first N ₀ bits are punctured at the beginning of the code block; the following must apply:

N _0. (E _minus - e _plus ) <e _ini ≤ N ₀ .e _minus - (N ₀ - 1) .e _plus (1).

Another criterion is that the last bit of the block is also punctured, according to the following condition:

(N ₀ - 1). (E _minus - e _plus ) <e _ini (2).

In this case, the value of the error variable e become negative at the last bit, which means that then this bit is punctured.

Both criteria are e.g. B. fulfilled by the following preferred choice of parameters:

e _plus = X _i - N ₀ (3),

e _minus = N _p - N ₀ (4),

e _ini = N ₀ .e _minus - (N ₀ - 1) .e _plus (5).

These formulas also include the special case that no bit should be punctured at the beginning of the code block (N ₀ = 0). The following then applies: e _ini = X _i , e _plus = X _i , e _minus = N _p .

The general implementations that choose e _ini according to the formulas (1) to (4), yield rate matching patterns which are different from those of the preferred choice of parameters according to (3) to (5) only in that from the (N ₀ + 1) -th to the (N _p - 1) -th puncturing point the index of the bit to be punctured can be decreased by 1.

For the application example of puncturing from 48 bits to 40 bits, the table in FIG. 26 shows puncturing patterns according to the preferred parameter selection up to N ₀ = 6. The puncturing points that are not printed in bold can be varied by varying the e _ini value according to (1) and (2 ) either partially or completely decreased by 1.

The table shown below in FIG. 27 shows in the same way the resulting patterns for puncturing from 111 bits to 80 bits.

Although this is not the best Dotting patterns discussed earlier were achieved, you can with this procedure but a certain improvement in the transmission quality compared to the current status of the specification the changes to be made are comparative are low.

repetition

According to the invention, the parameters of the rate matching algorithm are calculated so that a maximum distance of the last bit to be repeated from the end of the block is guaranteed, so the following must apply:

e _ini = 1 + X _i .e _minus - N _p .e _plus (6).

Furthermore, the average distance between bits R _R to be repeated can be specified. R _R need not be an integer, but can be a positive, rational number. The following then applies:

Thus, e _plus and e _minus can be freely chosen under the boundary condition that their quotient gives R _R and a total of N _p bits are repeated.

If the first bit to be repeated, more precisely the position of the first bit to be repeated (here referred to as b ₁ ), is to be specified, the following must also apply in addition to (6):


where e _minus should be an integer and b ₁ ≤ X _i - N _p + 1.

A preferred choice of parameters results in:

e _minus = N _p , (9),

e _plus = X _i - b ₁ + 1 (10),

e _ini = (b ₁ - 1) .N _p + 1 (11).

With this choice of parameters, the position of the first bit to be repeated is b ₁ and, as required, N _p bits are repeated.

Again, the repetition patterns that arise are not optimal compared to the patterns already discussed above. Nevertheless, a certain improvement in the transmission quality compared to the current state of the specification can be achieved with this method, the changes to be made again being comparatively small. A favorable choice of parameter b ₁ can ensure that the repetition does not start at the beginning. Repetition is not necessary at the beginning, since the bits at the beginning of the convolution decoder already have a comparatively low error rate as shown above. So it is much more useful if the bits to be repeated, as is done by this method, are concentrated more towards the middle. A disadvantage of this exemplary embodiment, however, is that it only avoids repetition at the beginning, while the conditions can be influenced far less positively at the end. This is the price that has to be paid for the simplified implementation.

Of course, a combination from above is also possible criteria mentioned when selecting a puncturing pattern possible. For example, you can have a pattern of two here Combine patterns presented by beginning at the beginning one pattern is used and at the end the end of the second pattern. Furthermore, it is irrelevant if the Bits are output in a changed order and at the same time the puncturing pattern is adjusted analogously. For example, the order of the polynomials in the Swap the convolutional encoder.

Claims (17)

1. Method for adapting the data rate of a data stream in a communication device, the data stream can be subdivided into at least one data block which contains transmission bits to be transmitted, the transmission bits are formed by a coding process from information-carrying input bits, in which certain transmission bits are removed (punctured) from a data block of the data stream in order to adapt the data rate, - whereby a puncturing pattern specifies which transmission bits are to be removed - And the puncturing pattern is designed such that preferably transmission bits are removed that depend on a few input bits via the coding process. 2. The method of claim 1, wherein the puncturing pattern is formed by the following steps: Determination of a cumulative puncturing strength which indicates which portion of information bits has been removed from the data block by removing transmission bits, Formation of a decision function depending on the cumulative puncturing strength, - Minimize the decision function to determine the puncturing pattern. 3. The method according to any one of the preceding claims, at which the puncturing pattern has over a puncturing rate Distance between transmission bits to be removed specifies, the puncturing rate for different Areas in the data block are different. 4. The method of claim 3, wherein the puncturing rate essentially in the middle of the data block equidistant distances between the bits to be removed having. 5. The method according to any one of the preceding claims, characterized, that the puncturing pattern is designed such that by dotting, from the front end of the to punctured data blocks viewed here, a section from the following series (bit positions) contains: 1, 4, 2, 3, 8, 7, 5, 6, 15, 12, 14, 11, 10, 9, where "1" is the first Corresponds to bit position. 6. The method according to any one of the preceding claims, characterized, that the puncturing pattern is designed such that by dotting, from the back end of the to punctured data blocks viewed here, a section from the following series (bit positions) contains: 0, 4, 6, 1, 2, 15, 12, 10, 9, 7, 4, 5, 18, 13, 8, where "0" is the corresponds to the last bit position. 7. The method according to any one of the preceding claims, characterized, that the puncturing pattern is designed such that 8 of 48 bits are punctured, namely bits 1, 2, 4, 8, 42, 45, 47, 48. 8. The method according to any one of the preceding claims, characterized, that the puncturing pattern is designed such that 31 of 111 bits are punctured, namely bits 1, 2, 3, 4, 5, 6, 7, 8, 12, 14, 15, 24, 42, 48, 54, 57, 60, 66, 69, 96, 99, 101, 102, 104, 105, 106, 107, 108, 109, 110, 111th 9. The method according to any one of the preceding claims, characterized, that the puncturing pattern is designed such that 14 of 54 bits are punctured, namely bits 1, 2, 3, 4, 7, 8, 36, 39, 42, 48, 51, 52, 53, 54. 10. The method according to any one of the preceding claims 1-8, characterized, that the puncturing pattern is designed such that 14 of 54 bits are punctured, namely bits 1, 2, 3, 4, 6, 7, 8, 39, 45, 48, 51, 52, 53, 54. 11. Method for adapting the data rate of a data stream in a communication device, the data stream can be subdivided into at least one data block which contains transmission bits to be transmitted, the transmission bits are formed by a coding process from information-carrying input bits, in which certain transmission bits are repeated (repeated) to adapt the data rate from a data block of the data stream, a repetition pattern is used to specify which transmission bits are to be repeated, - And the puncturing pattern is designed such that preferably transmission bits are repeated, which depend on many input bits via the coding process. 12. The method according to claim 11, wherein the repetition pattern is formed by the following steps: - Determination of a function of a cumulative repetition strength, which indicates what proportion of input bits was repeated by repeating transmission bits in the data block, - formation of a decision function depending on the cumulative repetition strength, - Maximize the decision function to determine the repetition pattern. 13. The method according to any one of the preceding claims 11 or 12 characterized, that the repetition rate of the repetition pattern, which specifies the distance between the bits to be repeated, essentially in the middle of the data block equidistant distances conditional and at the edge of the data block such large distances that no bits are repeated. 14. The method according to any one of the preceding claims 11 to 13 characterized, that the repetition pattern is designed such that 4 of 36 bits are repeated, namely bits 16, 18, 20, 22 15. The method according to any one of the preceding claims, characterized, that the data block in which the rate adjustment is performed, data encoded with a convolutional code includes. 16. Communication device with a rate adjustment device ( 6 ) for puncturing or repeating a data block of a data stream supplied to the rate adjustment device ( 6 ) in accordance with a specific rate adjustment pattern for adjusting the data rate of the data stream, the rate adjustment device removing puncturing or repeating bits corresponding to the rate adjustment pattern from the data block or repeated, characterized in that the rate adjustment device ( 6 ) is designed in such a way that it carries out the rate adjustment with a puncturing pattern or repetition pattern which is designed according to one of the preceding claims 1-15. 17. Communication device according to claim 16, characterized in that the communication device ( 1 ) is a mobile radio transmission or mobile radio reception device, in particular a UMTS mobile radio transmission or UMTS mobile radio reception device.