DE10101092B4 - Improvement of the bandwidth efficiency in a packet-oriented transmission with forward error correction - Google Patents

Abstract

Method for the packet-oriented transmission of data (D) from a transmitting unit (S) to a receiving unit (E) via a transmission medium having a data channel (DK) and a response channel (AK), comprising the following method steps:
Segmentation (1) of a data packet (DP) to be transmitted into a plurality of data segments (DS1 ... DSm),
Redundancy generation (R) by coding (2; C) of the data segments (DS1 ... DSm) with an error-detecting code and / or an error-correcting code,
- Transmission (3) of the coded data segments (DS1 ... DSn) from the transmitting unit (S) to the receiving unit (E) in several steps (n) by incrementally transmitting successive coded data segments via the data channel (DK),
characterized in that
from the transmitting unit (S) as long as coded data segments (DS1 ... DSn) with additional redundancy (R) are sent incrementally until the transmitting unit (S) either via the response channel (AK) a message (ACK) on the successful error-free decoding ( 4, DC) of the data packet (DP) to be transmitted by the receiving unit (E) or until all the redundancy (R) of the coded ...

Description

• – Segmentierung eines zu übertragenden Datenpakets in mehrere Datensegmente,
• – Redundanzgenerierung durch Codierung der Datensegmente mit einem fehlererkennenden Code und/oder einem fehlerkorrigierenden Code,
• – Übertragung der codierten Datensegmente von der Sendeeinheit zur Empfangseinheit in mehreren Schritten durch inkrementelles Senden aufeinanderfolgender codierter Datensegmente über den Datenkanal.

The invention relates to a method for the packet-oriented transmission of data from a transmitting unit to a receiving unit via a transmission medium having a data channel and a response channel with the following method steps:

• Segmentation of a data packet to be transmitted into several data segments,
• Redundancy generation by coding the data segments with an error-detecting code and / or an error-correcting code,
• - Transmission of the coded data segments from the transmitting unit to the receiving unit in several steps by incrementally transmitting successive coded data segments over the data channel.

at the packet-oriented transmission Data is generally available for a variety of reasons Packet loss. In packet-switched fixed networks, packet losses occur especially by overload caused at network nodes. This happens, for example, in succession a buffer overflow and queue overflow in data packet mediators or routers. A router evaluates in the Data packets contained and determined addressing information using routing tables the cheapest further Path. Based on this information it is decided what to do with the data packet continues to happen.

at mobile transmission channels (e.g. packet switched modes in Global System for Mobile Communication GSM, General Packet Radio Service GPRS or Universal Mobile Telecommunication System UMTS) are transmitted Data due to atmospheric disorders and multipath propagation strongly bothers. Through the multipath propagation a transmitted pulse generates several echoes. By the overlay several such signal components or multipath components with respective phase shift relative to the direct propagation path it comes to extinctions, which cause the so-called fast fading (almost fading). These and other multipath fades to lead sometimes considerable fluctuations in the reception field strength. Next frequency-selective fading occurs also a temporal Dispersion on, with the result that at the receiver individual symbols of the transmitted Data about sometimes several symbol durations "blur" and bit errors digital transmission data occur.

The currently most widespread transmission protocols at the transport level (layer or layer 4 of the OSI reference model), namely the User Datagram Protcol UDP and the Transmission Control Protocol TCP respond very critically to such bit errors in received data packets.

TCP is a connection-oriented transport protocol that is a logical Full-duplex point-to-point connection. It makes sure that data is accurate and in the desired order over one transmit the underlying IP network become. It extends the underlying IP (Internet Protocol) for functions for data backup and connection control.

As soon as a bit error within a packet by an error-detecting one Code is detected, the affected packet is discarded or erased (erasure). In other words, packet loss also occurs. So-called Reliable transport protocols such as TCP use repeating mechanisms (Automatic Repeat Request method) to get lost or discarded packets despite disturbed channels safe to the recipient to be able to transport. Erroneous or lost data will be redone at the sender requested. However, this is a very significant increase the transmission delay connected, but that can not be tolerated by many applications (e.g., communications applications).

Here Therefore, unreliable transport protocols such as e.g. UDP used. Compared to much more common TCP does not use UDP for error detection and correction. UDP works faster and has a smaller header, why the ratio from the number of user data (in bits) to data packet length (in bits) is better. UDP is more suitable for applications that send short messages and if necessary, they can repeat completely, or in applications that done in real time Need to become (Voice or video transmission).

The overall error correction therefore occurs within the application. This not only affects bit errors, but also the total loss of Data packets, because routers at high network load UDP datagrams immediately discard. The application can be used in special applications, e.g. in the Real-time range, in error detection and error correction of support additional, higher-layer protocols e.g. the RTP (Real-time Protocol). The basic principle of RTP is the use of Forward Error Control. This is enhanced by an extended Header allows in the additional information stand. These are e.g. the type of transferred User data (language, image data, etc.) or the time of creation the data, making the data easier to get into a specific correct Order brought or discarded after a certain time can be.

Due to the necessity of the complete Repetition of messages in case of errors requires the use of UDP in a network, however, even greater bandwidths than TCP and requires a more complicated bandwidth management of the network operator.

UDP alone thus causes no increase in transmission delay or Delay, offers for it but also no mechanisms to compensate for packet losses. Therefore, there are intensive efforts, forward error correction methods (Forward Error Correction, FEC) and packet-oriented transmission with unreliable transport protocols such as e.g. Connect UDP.

Ideally The resulting methods should adapt to the characteristics of the transmission channel adjust to not need more bandwidth than absolutely necessary is. This is especially true when transmitting over mobile channels, since there the Resource bandwidth particularly valuable, i. very expensive for the subscriber is.

• – Stop-And-Wait-ARQ bei zeichenorientierten Protokollen, bei denen nach jedem übertragenen Zeichen auf dessen Quittierung gewartet wird,
• – Continous-ARQ bei bitorientierten Protokollen, bei denen die Quittierung durch einen Fenstermechanismus geregelt wird, so dass bei richtiger Dimensionierung der Fenstergröße kontinuierlich gesendet werden kann (z.B. bei High Level Data Link Control HDLC, ein bitorientiertes synchrones Steuerungsprotokoll auf der Schicht 2 des OSI-Referenzmodells).

Conventionally, this adaptation of the error protection to the transmission channel is realized mainly by the use of hybrid ARQ methods. ARQ stands for Automatic Repeat Request and represents a method for data transmission in which a transmitting unit sends data and waits for its correct acknowledgment by the receiving unit. If necessary, the transmitting unit sends the data again. This process is repeated until the data has been transferred correctly. Essentially two methods are distinguished:

• - Stop-and-wait ARQ for character-oriented protocols, which wait for acknowledgment after each transmitted character,
• - Continuous-ARQ for bit-oriented protocols, in which the acknowledgment is controlled by a window mechanism so that, given correct sizing, the window size can be transmitted continuously (eg high-level data link control HDLC, a bit-oriented synchronous control protocol on the layer 2 of the OSI reference model).

there becomes the to be transferred Information first as part of a channel coding with an error-detecting code coded. A well-known representative of such error-detecting codes is CRC (Cyclic Redundancy Check). These are certain Group of algorithms that consist of a bit pattern after a given mathematical rule form a control bit pattern (usually 16 or 32 bits long) and add this additionally to the user data. Then done a coding with a very powerful error-correcting Code, i. a lot of redundancy is added.

The transfer the coded sequence over the disturbed Channel is then incrementally in several steps. First, will only a part of the coded data is transmitted. The receiver unit leads one Decoding based on the data available to him and evaluated using the error-detecting code, whether the amount of data received for a flawless Reconstruction is sufficient or not.

is error-free decoding is possible this the transmitting unit via a return channel communicated (e.g. an acknowledgment or confirmation, too referred to as acknowledgment ACK) and the transmission of the next information packet is prepared. If, however, an error-free decoding is not possible, requests the receiving unit the return channel additional Redundancy (often designated with NACK) and undertakes after their arrival one again decoding attempt, in which the receiving unit already received data combined with the newly added redundancy. It will be additional as long Redundancy requested until an error-free decoding is possible. A detailed Description of hybrid ARQ methods can be found in Lin, S./Costello, D .: "Error Control Coding: Fundamentals and Applications ", Prentice-Hall Inc., Englewood Cliffs, N.J., 1983.

Of the The advantage of this method is that actually only so much redundancy is transmitted will, as indeed needed becomes. However, this known method also has packet loss channels serious disadvantages.

So must the return channel a secure transmission guarantee, i. to have to ARQ procedures are also used there. This requires the implementation complex protocol stacks (e.g., timers). Dependent on the number of repetitions and the round-trip delay (the Time between the sending of information by the transmitting unit and the local arrival of a reaction of the receiving unit via the return channel) is currently in mobile transmission scenarios with very high delays to count. In multimedia applications, such as Streaming video, this means that in the receiver a very big one Buffer must be present. This also leads to increased complexity in the receiving terminal.

From WO 99/22481 a method and an arrangement for the transmission of data blocks is known, which are transmitted without error coding. In this case, if such a data block is received without error, simply the next block is transmitted, while if a data block is received with errors, a respective receiving station sends a message indicating an error Request for correction information. Following this, the data block source transmits the requested error correction information until the receiving station can decode the data block without error, the error correction information transmitted for this steping further adding redundancies until the receiving station is able to perform the correction. If the relevant data block has now been received without error, the method continues with the transmission of the next block.

It is known in point-to-multipoint transmissions data packets Encode message and send it with additional redundancy, and by means of a plurality of positive or negative answers to decide if and how the transfer the message continues or aborts and repeats should.

The The object of the present disclosure is therefore to the transfer across from to simplify the prior art and yet a Adaptation to the transmission channel to achieve.

According to the present Invention solves this problem by the method described in the preamble by is further developed that coded by the transmitting unit as long Data segments with additional Redundancy be sent incrementally until the transmitting unit either over the Answer channel a message about the successful error-free decoding of the data packet to be transmitted receives the receiving unit or until all Redundancy of coded data segments was transmitted.

To a first advantageous embodiment of the method according to the invention the coding is done in such a way that to be transmitted payload of a Data packets are divided into consecutive data segments of equal length which are around the respective redundancy symbols of an error-correcting Codes are added to respective codewords, where the code words for transmission be arranged so that each codeword is a column of all to be transmitted sequentially forms coded data segments.

there it has proven to be advantageous if coded first such Transfer data segments Be that exclusively User data of the transferred Contain data packets before transmitting encoded data segments which contain redundancy information.

The underlying principle of the invention will be briefly outlined in the following become. The in a plurality of information packets to be transmitted or Data packets present data are transmitted packet by data packet. In doing so, each data packet is first segmented into individual data segments and also with an error-detecting and subsequently with a powerful error-correcting code coded. Then the transmission begins also with the incremental transmission of the coded segmented Dates. Once the receiving unit, the information packet or data packet decode error-free on the basis of the already received data segments , it shares this with the transmitting unit via the response channel in the form of a confirmation or acknowledgment ACK with. receives the sending unit sends the ACK message, this stops the current transmission and starts the transmission the next Data packet. If the ACK message is lost on the answer channel or if there is no answer channel, the transmitting unit sets the transmission until all redundancy has been sent and begins then with the transfer a new data package.

in the Compared to known hybrid ARQ methods, the complexity of a Receiving unit can be significantly reduced. However, it may too a total loss of information, if the error protection used not high enough becomes. However, this can be largely avoided if the error-correcting code corresponding to the long-term characteristic of the transmission channel, especially the data channel, and the QoS parameter (Quality of Servie, QoS) of the bearer service.

Around the possibility a total loss of information, therefore, after another advantageous embodiment of the method of the invention of the error-correcting Code based on information of a long-term characteristic of the data channel chosen so that even with the worst possible transmission quality of the data channel there is sufficient error protection.

This For example, using the Real-time Control Protocol (RTCP) achieved as part of RTP various statistics on the current Evaluates connection (error rate, delay, etc.) and thus a check of QoS allows (not However, a high QoS allows itself).

A particularly advantageous embodiment The invention uses a Reed-Solomon code as the error-correcting code one. Examples of such error-correcting codes are the above-mentioned work by Lin, S./Costello, D. to take.

To The invention described above can be achieved of the cited Advantages also create a transmitting unit or a receiving unit, the above appropriate means of implementation of the inventive method for packet-oriented transmission of data.

Especially Advantageously, such transmission units and receiving units can be in communication terminals or communication networks.

there The best results are achieved if for a communication network transmission channels with as possible low round-trip time are provided.

One decisive advantage is that, in contrast to conventional ARQ method only a single response (ACK) needed is and not several. Furthermore can this answer about a little bit complex reverse link or response channel are transmitted. For the answer channel Therefore, an additional ARQ procedure is not necessary.

Further Details and advantages of the invention will become apparent from a in following illustrated advantageous embodiment and in conjunction with the figures. Here are elements with the same functionality with the the same reference numerals. It shows in schematic representation:

1 a transmitting and receiving unit connected via a faulty transmission channel,

2 a block diagram of a transmitting unit with the transmitter-side measures of the invention and

3 a block diagram of a receiving unit with the receiving side measures of the invention.

The principle of the invention will now be clarified with reference to an embodiment. The representation according to 1 shows the scenario of a transmission between a transmitting unit S and a receiving unit E via a disturbed transmission channel (indicated by a lightning-shaped arrow) with a data channel DK and a response channel AK. In this case, the response channel AK can be made technically simpler than the data channel DK, which is indicated by the thicker arrow DK compared to AK.

considered shall be an example of a streaming multimedia application on a mobile packet loss channel DK (i.e., the remote IP scenario). The QoS parameters of the 'packet switched bearer service' are known from the max. pack erasure / loss rate. Thus the required error protection for the most unfavorable all cases, i.e. a transmission at the max. packet erasure / loss rate. Thus, can always a faultless decoding DC can be achieved.

First, to be performed on the part of the transmitting unit S measures of the invention with reference to the in 2 illustrated schematic representation of a transmitting unit S are explained. The multimedia information D to be transmitted at specific intervals in the form of individual information packets is transmitted to data packet DP for data packet DP. The data packet DP to be transmitted currently is addressed, for example, by a pointer P pointing to this data packet. For transmission, each selected data packet DP is first segmented 1 and encoded with an error correcting systematic code designed for the worst case transmission case 2 ,

In The present example is the information block or the data packet DP with the payload N in the payload part of m consecutive transmission packets same length, hereinafter referred to as data segments DS1 to DSm distributed. Payload is generally referred to as the end user to disposal standing payloads to be paid by him in data packets, the paid become.

The Length of the Payload part is given as l symbols (e.g., bytes). In the payload part of the further data segments DSm + 1 to DSn are the redundancy symbols R of the error-correcting code, in this case, for example a Reed-Solomon (RS) code over Gabi's field GF (2 '\ 8), contain. The coding should be done in such a way that the entire transmission block (n data segments DS1 to DSn with payload length l) from exactly l RS codewords CW1 to CWl of length n bytes.

The i-th codeword CW in the transmission block is thus from top to bottom in 2 from the i-th byte of the payload part of each data segment DS1 to DSn together, is therefore systematic (1 = <i = <1).

The reason for this special type of coding is as follows: If a data segment DS1 to DSn of the transmission block is lost, then in all 1 code words CW1 to CW1 only one single symbol (here: 1 byte) is deleted at the same (known) position. Using well-known 'erasure decoding algorithms' can now use this symbol in each code word CW1 to CWl be reconstructed separately by means of the attached redundancy R, and thus ultimately the content of the entire transmission block DS1 to DSn or DP. Due to the special nature of RS codes, which are known to be 'maximum-distance-separable', this type of reconstruction works as long as the number of lost data segments is less than or equal to the number of redundancy segments R and DSm + 1 to DSn (in this Case so nm) is. This explanation may be sufficient to understand the following procedures.

After the addition of redundancy R by the coding 2 or C as described above, the data segments DS1 to DSn are transmitted in accordance with their order via the data channel DK 3 (from the point of view of 2 "from top to bottom"). This advantageously takes place in such a way that first the data segments DS1 to DSm, which contain user data N, and only then such data segments DSm + 1 to DSn with redundancy R are transmitted. This can be done, for example, by means of a shift register SH, in which a data segment is loaded and then output sequentially byte by byte via the data channel DK until all l bytes of a data segment have been output. Subsequently, the next data segment is loaded into the shift register and so on.

In doing so, packets or data segments are lost according to the statistical properties of the transmission channel DK. Since the mobile channel is time-variant, there are time periods in which the packet loss rate is much lower than that for the worst case. Then the in 3 also shown as a schematic diagram receiver unit E not all data segments DS1 to DSn for error-free reconstruction of the multimedia data to be transmitted data packet DP, but only a small proportion. The reception can also take place via a shift register SH, in which the incoming symbols are read sequentially and then as a complete data segment for decoding 4 or DC are output.

In the present example with the RS-coding 2 or C, the content of the information packets or data packets DP can be completely reconstructed as soon as at least m data segments of the n available data segments (no matter which) of the transmission block have been successfully received.

In the following it is assumed that due to an error burst (eg as a result of a fading hole on the mobile channel) the second and third data segments DS2 and DS3 are lost, while all other data segments are transmitted without errors. After receiving the first two redundancy segments DSm + 1 and DSm + 2, the receiver unit E is for decoding 4 or DC minimally required m data segments before.

After successful decoding DC of the payload N of a data packet DP now sends the receiver unit E via the response channel AK an ACK message to the transmitting unit S, due to the delay on the channel AK (expressed by the so-called round trip time) at the Sending unit S, for example, arrives at the time when it just wants to send the (nx) th data segment (with x <nm) in the sequence of data segments (ie the (nxm) -th redundancy packet). The transfer 3 is immediately stopped at this point, and the transmission of the next data packet DP or transmission block can be started by the pointer P is incremented and then addressed the next data packet DP, which is handled in the same way. The savings in this case are x full data segments.

In the case where n = 19 and m = 7 and the transmitting unit S is currently transmitting the 12th data segment when the acknowledgment ACK arrives (ie the 5th redundancy segment), the saving amounts to 8th Data segments. It is thus as little bandwidth wasted, and the less, the shorter the round-trip time of the channel DK and AK is.

An adaptation of the error protection to the transmission channel DK with AK is thus relatively easy even during the transfer feasible (so-called adaptive error protection). If the ACK message is lost, the no longer required x redundancy segments are also sent. These are discarded in this case, when they reach the receiver unit E, there. Then, although more redundancy R than required is transmitted, however, the terminal complexity is much lower than in the known hybrid ARQ method, which is why there are cost advantages over the known prior art. Since the ACK message is usually very small (a few few bytes), at the level of the physical layer (layer 1 of the OSI reference model) a very powerful error protection can be used to minimize the likelihood of loss.

Claims (10)

Method for the packet-oriented transmission of data (D) from a transmitting unit (S) to a receiving unit (E) via a transmission medium having a data channel (DK) and a response channel (AK), comprising the following method steps: 1 ) of a data packet (DP) to be transmitted into a plurality of data segments (DS1 ... DSm), - Redundancy generation (R) by coding ( 2 ; C) the data segments (DS1 ... DSm) with an error-detecting code and / or an error-correcting code, - transmission ( 3 ) of the coded data segments (DS1 ... DSn) from the transmitting unit (S) to the receiving unit (E) in several steps (n) by incrementally transmitting successive coded data segments via the data channel (DK), characterized in that the transmitting unit (S ) as long as encoded data segments (DS1 ... DSn) with additional redundancy (R) are sent incrementally until the transmitting unit (S) either via the response channel (AK) a message (ACK) about the successful error-free decoding ( 4 ; DC) of the data packet (DP) to be transmitted by the receiving unit (E) or until all redundancy (R) of the coded data segments (DS1 ... DSn) has been transmitted. Method for the packet-oriented transmission of data according to Claim 1, characterized in that the coding ( 2 ) in such a way that payload data (N) of a data packet (DP) to be transmitted is subdivided into successive data segments (DS1 ... DSm) of equal length (l) which are connected to the respective redundancy symbols (R) of an error-correcting code to respective codewords (CW1 ... CWl) are added, wherein the codewords (CW1 ... CWl) are arranged for transmission such that each codeword (CW1 ... CWl) one column of all encoded data segments to be transmitted sequentially (DS1 ... DSn ). Method for packet-oriented transmission of data as claimed 1 or 2, characterized in that first such encoded data segments (DS1 ... DSm) Be that exclusively User data of the transferred Contain data packets before encoded data segments (DSm + 1 ... DSn) which contain redundancy information (R). Method for packet-oriented transmission of data as claimed 1, 2 or 3, characterized in that the error-correcting one Code based on information of a long-term characteristic of the data channel (DK) so chosen becomes, that even with the worst possible transmission quality of the data channel (DK) there is sufficient error protection. Method for packet-oriented transmission of data after a of the preceding claims, characterized in that the error-correcting code is a Reed-Solomon code (RS) is used. Sending unit (S) for carrying out the method for packet-oriented transmission of data (D) according to any one of the preceding claims. Receiving unit (E) for carrying out the method for packet-oriented transmission of data (D) according to one of the preceding claims 1 to 5. Communication terminal with a transmitting unit (S) according to claim 6 and / or a receiving unit (E) according to claim 7th Communication network with at least one transmitting unit (S) according to claim 6 and a plurality of receiving units (E) according to claim 7. Communication network according to claim 9, characterized characterized in that transmission channels (DK, AK) with as possible low round trip time provided are.