BACKGROUND
[0001] Forms of communication and data transmission make use of low-density parity-check codes (LDPC). LDPC codes are a form of linear error correcting code, which are used to transmit messages over a noisy transmission channel.
[0002] When a transmitter uses an LDPC code to encode data, it encodes it in a redundant form. This redundancy allows the receiver to detect a limited number of errors that may occur anywhere in the message, and also allows the receiver to correct these errors without retransmission of data.
[0003] LDPC codes are powerful block codes whose performances get close to Shannon capacity as their block size increases. As a result, efficient and high performing LDPC codes have been standardized by several standardization bodies, the first being DVB-S2. In DVB-S2 two different coded block sizes have been used: 64800 and 16200 bits.
[0004] LPDC codes are finding increasing use in applications where reliable and highly efficient information transfer over bandwidth in the presence of data-corrupting noise. LPDC codes are currently utilized in digital video broadcasting, microwave communications, cell phone communications, and Wi-Fi.
[0005] FIG. 1 illustrates an example system
100.
[0006] As illustrated in the figure system
100 includes a gateway
102, a user terminal
104, a user terminal
106, a user terminal
108, a user terminal
110, a user terminal
112, and the internet
114.
[0007] User terminal
104 is arranged to send code blocks, which are 120 bits in length, to gateway
102 along bi-directional transmission line
116.
[0008] User terminal
106 is arranged to send code blocks, which are 240 bits in length, to gateway 102 along bi-directional transmission line
118.
[0009] User terminal
108 is arranged to send code blocks, which are 360 bits in length, to gateway
102 along bi-directional transmission line
120.
[0010] User terminal
110 is arranged to send code blocks, which are 480 bits in length, to gateway
102 along bi-directional transmission line
122.
[0011] User terminal
112 is arranged to send code blocks, which are 600 bits in length, to gateway
102 along bi-directional transmission line
124.
[0012] For purposes of discussion, in this example, each user terminal is transmitting discrete code blocks of a certain length. In practice, each user terminal is operable to transmit a plurality of code block lengths, which are dependent on the amount of information to be transmitted.
[0013] Gateway
102 is arranged to send and receive data from, bi-directional transmission line
116 from user terminal
104, bi-directional transmission line
118 from user terminal
106, bi-directional transmission line
120 from user terminal
108, bi-directional transmission line
122 from user terminal
110, and bi-directional transmission line
124 from user terminal
112. Gateway
102 is additionally arranged to send encoded data stream
126 to internet
114.
[0014] Gateway
102 contains a LDPC code and architecture for each respective code block length that it supports. In this example, gateway
102 contains an LDPC code and architecture set for each of the code block it supports, the lengths of which are 120 bits, 240 bits, 360 bits, 480 bits, and 600 bits.
[0015] In operation, a user will send code blocks to gateway
102 from a user terminal. Each individual user terminal encodes the data with an LDPC encoder and then sends the encoded data blocks to gateway
102.
[0016] When gateway
102 receives the encoded LDPC code blocks, it decodes them using an LDPC decoder to obtain the originally transmitted data. Note that when gateway
102 wants to send data to a user terminal the process is reversed.
[0017] When gateway
102 wants to send data to a user terminal, it uses an LDPC encoder to encode the data. Once the data is encoded, the encoded code blocks are transmitted to a user terminal, which then decodes the code blocks to obtain the originally transmitted data. Once the data is encoded, it may be sent to internet
112 via internet connection
126.
[0018] Data being sent over a noisy channel needs to be sent in a secure form. When a transmission channel is noisy it is much easier to lose data during transmission. This can create difficulties because of transmission control protocol (TCP).
[0019] Each data packet that is sent has a sequence number that is used to reconstruct the transmitted data in the correct order. For successful communication, the reliability of each packet (i.e. robustness against the channel noise) needs to be similar.
[0020] Another difficulty to consider during data transmission is that of power consumption. The larger a data packet is, the less energy per bit is consumed during transmission. This is due to the fact that the performance of modern error correcting codes, such as LDPC or turbo codes, improves with increasing block size. For very large data streams it becomes very important to try and minimize energy consumption.
[0021] For example, if a user wants to send a data stream that is 120,000 bits long, it would be much more efficient to send 20 packets that are 6000 bits in length than it would be to send 3000 packets that are 40 bits in length.
[0022] Data streams that are being transmitted first need to be configured into a data stream length that is supported be the receiver it is being sent to. This will be further illustrated by
FIGs. 2-4B.
[0023] FIG. 2 illustrates an example of various data stream lengths as well as supported data stream lengths of system
200.
[0024] As illustrated in the figure system
200 includes data stream sizes that are integer multiples of 120 bits. Data stream
202, data stream
206, data stream
212, data stream
222, and data stream 230 represent the data stream lengths that are supported by the system hardware.
[0025] Data stream
202 through data stream
230 represents all possible incoming data stream lengths. Note that it is possible of for an incoming data stream to exceed the possible supported data stream size but for the purposes of discussion there is a cut off at 1800 bits represented by data stream
230.
[0026] FIG. 3A and
FIG. 3B illustrates the transmission of a data stream size that is not supported by the receiver hardware.
[0027] As illustrated in
FIGs. 3A-3B system
300 includes supported data stream
222, data stream length
224, a data packet
326 and a data packet
328. Supported data stream length
222 is the length of 11 data packets, each of which is 120 bits in length. Data stream
224 is comprised of 12 individual data packets, data packet
302 through data packet
324, each of which is 120 bits in length.
[0028] Data packet
326 is a packet that can hold 11 data packets each of which are 120 bits in length. Data packet
328 is a packet that can hold 11 data packets each of which are 120 bits in length.
[0029] In operation, data stream
224 needs to be converted into a data stream, of two or more pieces, with lengths that can be put into data packets that are supported by the receiver hardware. Since data stream
224 is too long to fit into data packet
322, it needs to be divided up into smaller lengths that will fit into a supported data packet size that will then be transmitted.
[0030] The division of data stream
224 into smaller lengths to fit into supported data packet sizes will be further discussed in reference to
FIG. 3B.
[0031] FIG. 3B illustrates an example data stream that has been divided into smaller stream lengths that can fit into data packets which are supported by the receiver hardware.
[0032] In operation, data stream
224 has been divided into two lengths, which are small enough to fit into supported data packet
326 and data packet
328, both of which are the same length as supported data packet
222. Data packet
326 is filled with data packet
302 through data packet
322 from data stream
224 leaving data packet
324 left over.
[0033] Data packet
328 contains data packet
324 from data stream
224 but is otherwise empty. Since a data packet can't be sent unless it is full, the open space is filled with dummy bits. Dummy bits do not affect the content of data packet
328; they are simply placed in data packet
328 to fill it so it may be transmitted.
[0034] At this point, data stream
224 has been divided into two smaller lengths that have been put into data packet
326 and data packet
328, and they are ready to be transmitted to the receiver.
[0035] FIG. 4A and
FIG. 4B illustrate the transmission of a data stream length that is not supported by receiver hardware.
[0036] As illustrated in
FIGs. 4A-4B system
400 includes supported data stream length
206, data stream
224, a data packet
402, a data packet
404, a data packet
406, and a data packet
408. Supported data stream length
206 is the length of 3 data packets, each of which is 120 bits in length. Data stream
224 is comprised of 12 individual data packets, data packet
302 through data packet
324, each of which is 120 bits in length.
[0037] Data packet
402, data packet
404, data packet
406, and data packet
408 can each hold 3 packets of data, each of which are 120 bits in length.
[0038] In operation, data stream
224 needs to be converted into a data stream lengths that can be put into packet sizes that are supported by the receiver hardware. Since data stream
224 is too long to fit into data packet
402, it needs to be divided up into smaller lengths that will fit into a supported data packet size that will then be transmitted.
[0039] The division of data stream
224 into smaller lengths to fit into supported data packet sizes will further be discussed in reference to
FIG. 4B.
[0040] FIG. 4B illustrates an example data stream that has been divided into smaller stream lengths which are able to fit into data packet which are supported by the receiver hardware.
[0041] In operation, data stream
224 has been divided into 4 equal lengths, each of which are small enough to fit into data packet
402, data packet
404, data packet
406, or data packet
408. Each data packet is filled with information from data stream
224.
[0042] Once data packet
402, data packet
404, data packet
406, and data packet
408 are filled with information from data stream
224, they are ready to be transmitted to the receiver.
[0043] There are several problems with the current methods when utilizing LDPC codes. The first problem is that of large block size support as shown in
FIGs. 3A-4B. A code block may sometimes need to be very small and sometimes very large.
[0044] If only a large block size was implemented, a lot of transmission capacity would be wasted when the user had only a small amount of data to send. Conversely, if only a small block size was implemented, more power would be used than necessary when the user had a large amount of data to send since the performance of iterative decoding systems improve as the block size increases.
[0045] Even though a user may have an arbitrarily large amount of data to send occasionally, it would not be practically feasible to implement arbitrarily large block sizes. That's because the complexity of iterative receivers increases almost linearly with increasing block sizes.
[0046] Another problem with the current methods of utilizing LDPC codes, is the lack of intermediate block size support as shown in
FIG. 1. If one wants to support many different code block lengths that it may be undesirable to design a separate LDPC code for each block length.
[0047] The final problem with the current methods for utilizing LDPC codes as shown in
FIG. 1 is the space that is taken up by the architecture used for each LDPC code. If one wants to support many different code block lengths, an LDPC code and its respective architecture needs to be implemented in the hardware design for each individual code block length.
[0048] To support many different code block lengths, the architecture needed to support each individual code block length begins to be problematic. As the number of architectures being implemented for each code block length, the complexity of the hardware design increases as well.
[0049] The increase in complexity, for each supported code block length, is too high when attempting to design hardware that is capable of supporting many code block lengths.
[0050] What is needed is a system and method to address the problem of large block size support. The transmitted block sizes need to be chosen so that the minimum transmitted size is maximized. This means that when a user wants to transmit more data than the maximum supported block size, the transmitted block sizes should be as large as possible without having empty space in the transmitted block.
[0051] What is needed is a system and method to address the problem of intermediate block size support. To address intermediate block size support, bits in the code should be shortened, punctured, and repeated to derive intermediate code block lengths. On the other hand, the location of punctured bits should be carefully chosen in order not to cause substantial performance loss. What is needed is a system and method for using common decoder architecture for different code rates and block sizes. It may also be desirable to use the same piece of hardware to decode all block sizes and all code rates with the help of a code-specific set of parameters. This will reduce the complexity of the receiver where a different decoder is used for different codes.
BRIEF SUMMARY
[0052] The present invention provides a system and method for supporting large block sizes. By choosing to transmit incoming data blocks such that, the minimum transmitted size is maximized, one can transmit blocks reliably and efficiently despite the original code block being too large to be supported.
[0053] The present invention additionally provides a system and method for supporting intermediate block sizes. Through puncturing, shortening, and repeating incoming code blocks as indicated by a parity check matrix it is possible to create code blocks of an intermediate length.
[0054] The present invention additionally provides a system and method for using common decoder architecture for different code rates and block sizes. By using a hardware architecture that is common to all code blocks, it is possible to support many different codes by running them through an iterative process utilizing the architecture common to all code blocks.
[0055] In accordance with an aspect of the present invention, a method is provided for supporting code blocks of any length and converting them into code blocks that are large or intermediate in length, which are compatible with a system's hardware. A method is also provided for using a common decoder architecture for processing all incoming code blocks.
[0056] In accordance with an aspect of the present invention, an encoder is provided to encode a string of information bits having a bit length
k into a string of encoded bits having a bit length
N. The encoder includes a plurality of LDPC portions and a controller. Each of the LDPC portions can encode a portion of the information bits into a LDPC code block, respectively, wherein each of the respective LDPC code blocks has a different bit length. The controller can instruct a number of the LDPC portions to encode the information bits into a plurality of LDPC code blocks such that the minimum encoded block size is maximized.
[0057] In accordance with another aspect of the present invention, an encoder is provided to encode a string of information bits having a bit length
k into a string of encoded bits having a bit length N. The encoder includes a LDPC portion and a controller. The LDPC portion can encode information bits into LDPC code blocks, wherein each of the LDPC code blocks has a bit length
S1. When N <
S1, the LDPC portion can do at least one of pad, shorten, puncture or repeat a bit within one of the string of information bits and parity bits associated with the string of information bits to create an encoded LDPC code block having a length N.
[0058] In accordance with another aspect of the present invention, a decoder is provided for use with a first LDPC code block and a second LDPC code block. The first low LDPC block has a first bit length, and the second LDPC block has a second bit length, wherein the first bit length is shorter than the second bit length. The decoder includes a receiving portion, a memory portion an address table portion, a processing engine and a controller. The receiving portion can receive a code block. The memory portion can store the first low density parity check code block as a first data structure and can store the second low density parity check code block as a second data structure. The address table portion has a first address table stored therein and has a second address table stored therein. The first address table relates a structure of the first LDPC code block to the first data structure, whereas the second address table relates a structure of the second LDPC code block to the second data structure. The processing engine can process blocks of data having a third bit length, wherein the third bit length is shorter than the first bit length. When the code block is the first low density parity check code block, the controller can instruct the memory portion to store the first LDPC code block as the first data structure and can instruct the processing engine to process the blocks of data within the stored first LDPC code block based on the first address table. When the code block is the second LDPC code block, the controller can to instruct the memory portion to store the second LDPC code block as the second data structure and can instruct the processing engine to process the blocks of data within the stored second LDPC code block based on the second address table.
[0059] Additional advantages and novel features of the invention are set forth in part in the description which follows, and in part will become apparent to those skilled in the art upon examination of the following or may be learned by practice of the invention. The advantages of the invention may be realized and attained by means of the instrumentalities and combinations particularly pointed out in the appended claims.
BRIEF SUMMARY OF THE DRAWINGS
[0060] The accompanying drawings, which are incorporated in and form a part of the specification, illustrate an exemplary embodiment of the present invention and, together with the description, serve to explain the principles of the invention. In the drawings:
FIG. 1 illustrates an example gateway and user terminals for a prior art configuration;
FIG. 2 illustrates various data stream lengths as well as supported data stream lengths;
FIG. 3 illustrates the use of packets that are too large for a given data stream length for the prior art of FIG. 1;
FIG. 4 illustrates the division of a data stream into many smaller lengths for the prior art of FIG. 1;
FIG. 5 illustrates an example gateway and user terminal configuration, in accordance with aspects of the present invention;
FIG. 6 illustrates the shortening, puncturing, and repeating of a code block, in accordance with aspects of the present invention;
FIG. 7 illustrates the optimization of data stream division, in accordance with aspects of the present invention;
FIG. 8 illustrates the gateway operation of the configuration described in FIG. 5, in accordance with aspects of the present invention; and
FIG. 9 illustrates the use of common decoder architecture, in accordance with aspects of the present invention.
DETAILED DESCRIPTION
[0061] Aspects of the present invention provide a method for supporting large code block lengths. In contrast with the embodiment of
FIGs. 2-4B, code blocks that are larger than the supported code block length available are broken up into small pieces in such a way that the minimum transmitted size is maximized.
[0062] In contrast to the prior art system described above, where the code blocks were either broken up into many small pieces or pieces of random length, and were then sent out to a receiver. These two methods were unreliable and inefficient.
[0063] Now the code blocks are now broken up into the largest possible size that can be transmitted without having wasted space. The intelligent decision to choose the block size to be transmitted such that they are as large as possible without having empty space, optimizes reliability and efficiency.
[0064] Choosing large block sizes exploits the fact that power is more efficiently utilized when sending large block sizes. Additionally, this decision can provide the desired degree of reliability for successful transmission of code blocks.
[0065] Aspects of the present invention provide a method for supporting intermediate code block lengths. Practical LDPC codes of interest have structured parity check matrices, therefore their storage is relatively much easier than random parity check matrices. On the other hand, in some applications, the number of desired parity check matrices may be so large that designing a separate LDPC codes for each block size may be undesirable.
[0066] In these cases, it may be possible to design a certain number of mother parity check matrices and derive the intermediate codes by shortening, puncturing or repeating. Provided the number of shortened, punctured or repeated bits is not excessive with respect to the original block size, performance loss would not be significant by carefully choosing the location of the affected bits.
[0067] The general rule provided, in accordance with aspects of the present invention, states that for lower code rates such as 1/2, 2/3 and 4/5 punctured bits should be chosen among the parity bits, whereas for high code rates such as 8/9 and 9/10, punctured bits should be chosen among the systematic bits to minimize the performance loss.
[0068] Aspects of the present invention provide a method for using common decoder architecture for different code rates and block sizes. As previously discussed, in prior art system 100, the current method for supporting many different code block lengths adds an undesired amount of complexity to a system.
[0069] Currently, a separate LDPC code and piece of architecture is implemented to support each different code block length and code rate, while not problematic when supporting a limited number of code blocks, designs become extremely complex when attempting to support many code blocks.
[0070] The present invention provides a method for finding a common decoder architecture, and then using the architecture to support all the code blocks.
[0071] A fully parallel LDPC decoder implementation would have as many bit nodes as the number of coded bits and as many check nodes as the number of parity check equations. While this implementation would have very high speeds, it would be extremely complex. Moreover it may need to have as many instantiations of the decoder as the number of parity check matrices that would be implemented.
[0072] A more practical approach may be to implement a partially parallel decoder where a certain number
M of physical parallel engines would process
M logical nodes of the LDPC codes at a time. This approach not only simplifies the decoder significantly, but also allows to support all of the different parity check matrices using a common decoder hardware, provided the same inherent structure is kept among all of the codes. Only a code specific table would change whenever a new parity check matrix is deployed. This table would tell the decoder parallel engines where to access the relevant edges for the logical nodes being processed.
[0073] The number of parallel engines
M influence the decoder speed and complexity, and it is determined by the code design. When a set of different parity check matrices is supported with a common hardware, the number of physical nodes has to be as large as the maximum
M among all of the codes. Therefore in order not to waste the decoder area for certain codes,
M for different parity check matrices should be as close to one another as possible.
[0074] The first two aspects of the present invention will now be described in greater detail with reference to
FIGs. 5-8.
[0075] In accordance with the first aspect, an encoder is provided to encode a string of information bits having a bit length
k into a string of encoded bits having a bit length
N. The encoder includes a plurality of LDPC portions. Each LDPC portion can encode a portion of the information bits into a LDPC code block, respectively, wherein each of the respective LDPC code blocks has a different bit length. The encoder additionally includes a controller that can instruct a number of the LDPC portions to encode the information bits into a plurality of LDPC code blocks such that the minimum encoded block size is maximized.
[0076] For example, suppose for purposes of discussion that an encoder in accordance with second aspect of the present invention includes a controller that instructs an encoder to encode a string of information bits using an LDPC portion. For purposes of discussion, presume that in this example that the string of information bits has a length
k=6720 bits. If the largest LDPC portion is only 5760 bits, the string of information bits must be broken into two separate pieces. The controller will determine the optimum LDPC portion to be used to encode the information bits based on performance considerations.
[0077] In this example, the controller would instruct an LDPC portion that was 3360 bits in length to encode the string of information bits. This choice would minimize the amount of dummy bits that were needed to fill each code block while simultaneously minimizing the amount of power needed to transmit each bit after being encoded.
[0078] In accordance with the second aspect, an encoder is provided to encode a string of information bits having a bit length
k into a string of encoded bits having a bit length
N. The encoder includes an LDPC portion and a controller. The LDPC portion can encode the information bits into LDPC code blocks, each of which has a bit length
S1. When N <
S1, the LDPC portion can pad, shorten, puncture or repeat a bit within one of the string of information bits and parity bits associated with the string of information bits to create an encoded LDPC code block having a length
N.
[0079] For example, returning to
FIG. 2, suppose for purposes of discussion that an encoder in accordance with second aspect of the present invention includes a first LDPC portion that can encode information bits into LDPC code blocks corresponding to data stream length
222 (in this case data stream length
222 corresponds to
S1)
. In accordance with the present invention, if a data stream length corresponding to data stream length
220 (in this case data stream length
220 corresponds to
N, in which N <
S1) were needed to be encoded, then LDPC portion would be able to pad, shorten, puncture or repeat a bit within one of the string of information bits and parity bits associated with the string of information bits in data stream length
220 to create an encoded LDPC code block having a length corresponding to data stream length
222 (in this case data stream length
222 corresponds to
S1).
[0080] FIG. 5 illustrates an example system
500 in accordance with aspects of the present invention.
[0081] As illustrated in the figure system
500 includes a gateway
502, a user terminal
504, a user terminal
506, a user terminal
508, a user terminal
510, a user terminal
512, and the internet
514.
[0082] User terminal
504 is arranged to send code blocks, which are 960 bits in length, to gateway 502 along bi-directional transmission line
516.
[0083] User terminal
506 is arranged to send code blocks, which are 3360 bits in length, to gateway 502 along bi-directional transmission line
518.
[0084] User terminal
508 is arranged to send code blocks, which are 3600 bits in length, to gateway 502 along bi-directional transmission line
520.
[0085] User terminal
510 is arranged to send code blocks, which are 5760 bits in length, to gateway 502 along bi-directional transmission line
522.
[0086] User terminal
512 is arranged to send code blocks, which are 6720 bits in length, to gateway 502 along bi-directional transmission line
524.
[0087] For purposes of discussion, in this example embodiment, each user terminal is transmitting discrete code blocks of a certain length. In practice, each user terminal is operable to transmit a plurality of code block lengths, which are dependent on the amount of information to be transmitted.
[0088] Gateway
502 is arranged to send and receive data from, bi-directional transmission line
516 from user terminal
504, bi-directional transmission line
518 from user terminal
506, bi-directional transmission line
520 from user terminal
508, bi-directional transmission line
522 from user terminal
510, and bi-directional transmission line
524 from user terminal
512. Gateway
502 is additionally arranged to send encoded data stream
526 to internet
514.
[0089] In operation, a user will send code blocks of a certain length to gateway
502. Each individual user terminal encodes the data with an LDPC encoder and sends the encoded blocks to gateway
502.
[0090] Gateway
502 contains architecture for each respective mother code that it supports. In this example, gateway
502 contains architecture for mother codes, the lengths of which are 960 bits, 3600 bits, and 5760 bits. Code blocks transmitted by UT1
504, UT3
506, and UT4
510 are able to be transmitted easily due to the fact that the code blocks are of a length that corresponds to a supported mother parity check matrix, and do not need to undergo any manipulation.
[0091] Code blocks from UT2
506 and UT5
512 are not of a length that is readily usable by gateway
502 and therefore have to be either shorten, punctured and repeated or padded to a length that corresponds to a mother parity check matrix which is further described with reference to
FIG. 6.
[0092] When gateway
502 receives the encoded LDPC code blocks, it decodes them using an LDPC decoder that has a length that is common to all incoming code block lengths. The LDPC decoder runs an iterative process to decode each incoming code block of any length and obtain the originally transmitted data.
[0093] Note that when gateway
502 wants to send data to a user terminal the process is reversed. When gateway
502 wants to send data to a user terminal, it uses an LDPC encoder to encode the data. Once the data is encoded, the encoded code blocks are transmitted to a user terminal, which then decodes the code blocks to obtain the originally transmitted data.
[0094] FIG. 6 illustrates the shortening, puncturing, and repeating (SPR) of code block length to a length corresponding to a supported LDPC code by using a mother parity check matrix.
[0095] As illustrated in the figure, system
600 includes gateway
502, UT2
506, UT3
508, SPR portion
602, SPR portion
604.
[0096] UT2
506 is arranged to send data blocks, 3360 bits in length, to SPR portion
802. UT3
508 is arranged to send data blocks, 3600 bits in length, to SPR portion
604.
[0097] Gateway
502 is arranged to receive data blocks that are 3600 bits in length and to send data blocks 3600 bits in length to a receiver via internet connection
620.
[0098] In operation, a user using UT3
508 will send data to a receiver (not shown). The code blocks that UT3
508 sends out are 3600 bits in length, which is received by SPR portion
604. SPR portion
604 has a list of LDPC codes, the code block length of the supported LDPC codes, and their mother parity check matrices used by gateway
502. In this example, the supported LDPC code is associated with a code block whose bit length is 3600.
[0099] SPR portion
604 will receive the code block sent by UT3
508 and cross reference the code block length with code block lengths that are associated with the supported LDPC codes contained by gateway
502. In this case SPR portion
604 finds that the code block sent by UT3
508 is compatible with a LDPC code supported by gateway
502 and sends the code block to the gateway without any code block length manipulation.
[0100] Simultaneously, a user using UT2
506 will send data to a receiver (not shown). The code blocks that UT2
506 sends out are 3360 bits in length, which is received by SPR portion
602. SPR portion
602 has a list of code block sizes that are associated with LDPC codes supported by gateway
502. In this example, the supported block size is 3600 bits in length. The code block sent by UT2
506 is too small to be used by a mother LDPC code and the LDPC code for a code block 3360 bits in length must be derived from a supported mother parity check matrix by shortening, puncturing, and repeating.
[0101] When a user wants to transmit a code block of a length that is not compatible with a supported LDPC code, an LDPC code must be shortened, punctured, and repeated to be made usable. This means that a (
n2, k2) code with a code rate of
R=
k2/
n2 must be derived from a (
n1,k1) code with a code rate of
R=
k1/
n1.
[0102] To derive a (
n2,k2) code, a code block of k
2 bits must first be accepted. Once the code block has been received,
k1-k2 dummy bits must be added to the code block of length
k2 to obtain a code block of length
k1. At this time, the code block of length
k2 has been transformed into a code block of length
k1, which enables the use of the existing (
n1,k1) LDPC code to obtain
n1 coded bits.
[0103] Once the bits have been encoded, the
k1-k2 dummy bits that were previously added in must be removed from the
n1 coded bits. The number of coded bits left after the removal of the dummy bits is
n1-(
k1-k2) bits. Finally
n1-(
k1-k2)
-n2 bits must be punctured from the remaining coded bits to obtain
n2 bits. The process of deriving an LDPC code through shortening, puncturing, and repeating will now be described.
[0104] UT2
506 is sending code blocks of 3360 bits at a code rate of 1/2 (i.e., the ratio of information bits to coded bits is 1/2), so the LDPC code that is desired is denoted as (
6720/3360). Since the code block does not have a supported LDPC code, it needs to be derived using a mother parity check matrix of a LDPC code that is supported. In this case the existing LDPC code is denoted as (
7200/3600).
[0105] To begin deriving an LDPC code for a code block of 3360 bits and a rate of 1/2, the 3360 bits are accepted. 240 dummy bits are added to the 3360 information bits to obtain a code block that is 3600 bits in length. Using the supported LDPC code and associated parity check matrix, a code block of 7200 bits is derived.
[0106] The 240 dummy bits that were added to the original code block of 3360 bits are still contained within the 7200 bit code block. Removing the dummy bits a code block of 6960 bits is obtained. Since a code block of 6720 bits is desired, another 240 bits must be punctured to obtain 6720 bits as desired. After a code block of 6720 bits has been derived, SPR portion
602 will send the coded bits to gateway
502. Once the code has been received by gateway
502, the code can be reconstructed using the mother parity check matrix to recover the bits that were punctured out of the code block.
[0107] In slower code rates such as 1/2, 2/3, and 4/5 the punctured bits should be chosen among the parity bits, whereas for high code rates such as 8/9 and 9/10, punctured bits should be chosen among the systematic bits to minimize performance loss.
[0108] Shortening, puncturing, and repeating provides a method for coding and transmitting blocks of an intermediate size. This method allows the transmission of code blocks that may not be associated with a supported mother LDPC code.
[0109] In this example, the code block derived is 6720 bits in length, which is larger than the maximum implemented block size of 5720 bits. In this case, more than one code block needs to be sent to transmit all of the user's data. One method would be to transmit a block size of 5760 bits and another block size of 960 bits. This choice, would have poor performance due to the short block size of 960 bits because the shorter the block size the more transmit power per bit is needed. An optimum choice would be two 3360 bit blocks, because reliability and efficiency are optimized. The transmitted block sizes should be chosen such that the minimum transmitted size is maximized. This rule applies even if the number of transmitted block sizes is more than two. In the above example, the two 3360 bit blocks are sent by making use of the 3600 bit mother blocks.
[0110] The division of large code blocks in such a way that the minimum transmitted size is maximized will now be discussed further with respect to
FIGs. 7A-7B.
[0111] FIG. 7A and
FIG. 7B illustrate the transmission of a data stream length that is not supported by receiver hardware in accordance with aspects of the present invention.
[0112] As illustrated in
FIGs. 7A-7B system
700 includes supported data stream length
702, data stream
704, a data packet
706, and a data packet
708. Supported data stream length
702 is the length of 6 data packets, each of which is 560 bits in length. Data stream
704 is comprised of 12 individual data packets, data packet
710 through data packet
732, each of which is 560 bits in length.
[0113] Data packet
706 and data packet
708 can each hold 6 individual packets of data, each of which are 560 bits in length. Data packet
706 and data packet
708 are of the same length as data stream
702, the length of which is supported by the receiver hardware.
[0114] Occasionally a user may have an arbitrarily large amount of data to send, it would not be practically feasible to implement arbitrarily large block sizes. This is because the complexity of iterative receivers increases almost linearly with increasing block sizes. On the other hand, performance does not improve indefinitely with increasing block sizes; after every block size increase, the performance (reduction in transmit power per bit) reduces until a point where further block size increase causes negligible performance improvement. Hence, a system design was chosen based on performance and complexity considerations, leading to a maximum supported block size of, for example, 5760 bits.
[0115] In operation, data stream
704 is 6720 bits in length which is too large to fit into the largest supported data packet of 5720 bits. Therefore, data stream
704 needs to be converted into data stream lengths that can be put into packet sizes that are supported by the receiver hardware. Since data stream
704 is too long to fit into a supported data packet size, it needs to be divided into smaller lengths that will fit into a supported data packet size that will then be transmitted.
[0116] The division of data stream
704 into smaller lengths to fit into supported data packet sizes will further be discussed in reference to
FIG. 7B.
[0117] FIG. 7B illustrates an example data stream that has been divided into smaller stream lengths which are able to fit into data packets which are supported by the receiver hardware.
[0118] In operation, data stream
704 has been divided into two equal lengths, each of which is small enough to fit into data packet
706 and data packet
708. Data packet
786 and data packet
706 are filled with six individual data packets from data stream
704. Once data packet
706 and data packet
708 are filled with information from data stream
704, they are ready to be transmitted to the receiver.
[0119] In this configuration, the transmission of data stream
704 has been optimized in regards to reliability. With concerns relating to reliability, each data packet has equally and fully been filled with information from data stream
704. Elimination of the use of dummy bits provides the best and most reliable form of transmission of information.
[0120] Also in this configuration, the transmission of data stream
704 has been optimized in regards to energy consumption. With concerns relating to energy consumption, each data packet has the longest possible length without using dummy bits, because the larger the packets size, the more efficient the transmission.
[0121] Once the code blocks being transmitted by a user have been shorten, punctured, and repeated using a mother LDPC code and divided into lengths that are supported by the receiver hardware, gateway
502 is able to transmit the code blocks to its intended destination.
[0122] Note that when gateway
502 wants to send data to a user terminal the process is reversed. When gateway
502 wants to send data to a user terminal, it uses an LDPC encoder to encode the data. Once the data is encoded, the encoded code blocks are transmitted to a user terminal, which then decodes the code blocks to obtain the originally transmitted data.
[0123] FIG. 8A illustrates gateway
502 at time
t1 in accordance with aspects of the present invention.
[0124] As illustrated in the figure system
800 includes a receiver portion
802, a low density parity check portion (LDPC)
804, a controller
806, and a transmitter
808.
[0125] Receiver
802 is arranged to receive incoming data signal
810.
[0126] LDPC portion
804 is arranged to receive data via data signal
812 from receiver
802. LDPC portion
804 is additionally arranged to send controller alert
814 to controller
806 and receive enable signal
816. LDPC portion
804 is additionally arranged to send encoded data
818 to transmitter
808.
[0127] In operation, encoded data is sent to the encoder by a first user terminal (UT1)
822. UT1
822 reaches the encoder through incoming data signal
810. (Note that there may be a plurality of signals coming in from different user terminals.) At this point LDPC portion
802 decodes the encoded data.
[0128] UT
1 822 that is received by receiver
802 is sent to LDPC portion
804 via data signal
812. At this point LDPC portion
804 sends controller alert signal
814 to controller
806. Controller alert signal
814 tells controller
806 that LDPC portion
804 has received data and is ready to encode it.
[0129] Once alerted that LDPC portion
804 has data that needs to be encoded, controller
806 sends enable signal
816 to LDPC portion
804. Once the signal is received, LDPC portion
804 begins encoding data as described in
FIG. 6. Once LDPC portion
804 has encoded the data, it breaks the data up into pieces such that the minimum transmitted size is maximized, then the data is put into packets.
[0130] After the data is encoded and put into packets it becomes encoded first user terminal data (EUT1)
824. EUT
1 824 is sent to transmitter
808. At this point transmitter
808 sends EUT
1 824 to a receiver via internet connection
820.
[0131] FIG. 8B illustrates encoder
502 at time
t2 in accordance with aspects of the present invention.
[0132] In operation, encoder system
800 has already received UT1
822, encoded it, and transmitted it as EUT1
824. Now encoder
502 is receiving user terminal 2 data (UT2)
826 via incoming data signal
810.
[0133] Encoder
502 follows the same process as before, receiver
802 sends UT2
826 to LDPC
804 via data signal
812. Again, LDPC portion
804 alerts controller
806 via controller alert signal
814, that it has received data and is ready to encode it. Controller
806 then sends enable signal
816 to LDPC portion
804, allowing it to encode UT2
826 and then send it to transmitter
808 as encoded second user terminal data (EUT2)
828.
[0134] Again LDPC
804 encodes then breaks up the data into pieces such that the minimum transmitted size is maximized and then puts it into packets to be transmitted. After the data is encoded and put into packets it becomes encoded second user terminal data (EUT2)
828. EUT2
828 is sent to transmitter
808, where EUT2
828 is sent to a receiver via internet connection
820.
[0135] The third aspect of the present invention will now be described in greater detail with reference to
FIG. 9.
[0136] In accordance with the third aspect, a decoder is provided for use with a first low density parity check code block having a first bit length and a second low density parity check code block having a second bit length. The decoder may store the first low density parity check code block and then use a processing engine to process the stored low density parity check code block based on the address of the first code block. The decoder may store the second low density parity check code block and then use a processing engine to process the stored low density parity check code block based on the address of the second code block.
[0137] For example, suppose for purposes of discussion that a decoder in accordance with the third aspect of the present invention includes a first memory portion that contains a first LDPC code block and a second memory portion that contains a second LDPC code block. The decoder will decode the first and second code blocks by using a third LDPC code that is common to both the first and second code blocks.
[0138] FIG. 9 illustrates the use of common decoder architecture for different code rates and block sizes in accordance with aspects of the present invention.
[0139] As illustrated in the figure system
900 includes, a decoder
902, a LDPC portion
904, an address table
906, an address table
908, and a processor
910.
[0140] Decoder
902 is arranged to receive incoming encoded data from signal
912 and signal
914. Decoder
902 is operable to store incoming encoded data in either address table
906 or address table
908.
[0141] LDPC portion
904 is arranged to send decoding information to decoder
902.
[0142] Address table
906 and address table
908 are arranged to receive and store data sent by decoder
902.
[0143] Processor
910 is operable to process information stored in either address table
906 or address table
908. Processor
910 is additionally operable to send data to a receiver via signal
924.
[0144] In operation, encoded data is sent to decoder
902 from an outside source (not shown). The encoded data received will be in a length that is supported by the decoder. In this example the block are 120 bits in length and 600 bits in length, but in use incoming data may be in many different supported lengths.
[0145] Having architecture to support each different LPDC code takes up valuable space and adds an undesired amount of complexity to hardware design. When code blocks of different lengths are being received by the decoder it searches LPDC portion
904 for a LPDC code that is common among all of the incoming data blocks.
[0146] In this example, code block lengths that are supported by LDPC portion
904 are 22 bits and 40 bits in length. Since 22 is not a factor of 120 or 600 it cannot be used, but because 40 is a common factor of both 120 and 600, it can be used to decode the incoming encoded data. Note that in this example LDPC portion
904 only supports two code block lengths but in practice it may support many code block lengths.
[0147] Clearly there may not be a need to necessarily find a common factor of 120 and 600. For purposes of discussion, consider that a 120 bit block has an LDPC code rate of 4/5, such that it has 120 bit nodes and 24 check nodes. Some common factor of 120 and 24 should be found. Suppose the number 24 is chosen. In such a case, 24 hardware processing engines may be implemented that will process 120 bit nodes and 24 check nodes.
[0148] Now, consider that a 600 bit block has an LDPC code rate of 9/10, such that it has 600 bit nodes and 60 check nodes. In such a case, 60 hardware processing engines may be implemented to process this code. However, 60 hardware processing engines may not be a very good choice. In particular, once 60 processing engines are implemented in hardware, if an LDPC code rate of 4/5 is desired to be supported for a 120 bit block, then 36 of the 60 hardware processing engines would be idle.
[0149] To prevent this idle hardware processing engine problem, maybe 30 engines would be a better choice for an LDPC code rate of 9/10 for a 600 bit code, because then only 6 hardware processing engines would be idle for the LDPC code rate of 4/5 for a 120 bit block. Of course another point to consider is that as the number of engines increases, the decoder speed increases. Assuming that all the codes need a similar decoder speed, it is advantageous to choose the number of parallel engines similar for all codes, so that only a few engines stay idle in the hardware.
[0150] At this time decoder
902 sends the encoded data that is 120 bits in length to address table 906 and it sends encoded data that is 600 bits in length to address table
908. Code blocks of different lengths are sent to different address tables because a different matrix is used to decode each respective code block.
[0151] The parity check matrix used to encode code blocks 120 bits in length is also used to decode code blocks that are 120 bits in length. The parity check matrix used to encode code blocks 600 bits in length is also used to decode 600 bits in length. Segregating code blocks in to address tables for each length indicates to processor
910 which decoding matrix should be used.
[0152] Simultaneously decoder
902 sends the common LDPC code to processor
910 which will be used to decode the encoded information stored in address table
906 and address table
908.
[0153] Processing engine
910 takes encoded data from address table
906, which is 120 bits in length, and begins decoding the data. Since the LDPC decoding information is smaller than the block sizes stored, it must run the encoded data through an iterative process. With a 40 bit LDPC code, 120 bit code blocks must be run through the decoding process 3 times.
[0154] If processing engine
910 is decoding encoded data from address table
908, which is 600 bits in length, it must run the code blocks through an iterative process 15 times.
[0155] Once processing engine
910 has decoded the data from address table
906 and address table 908 it can send the data to a receiver (not shown) via signal
924.
[0156] As described above, using a common LDPC code to decode data of various lengths can be done while using common architecture. In previous methods, architecture was needed for each separate LDPC code, which added unwanted complexity to the hardware design.
[0157] System
500, in accordance with aspects of the present invention, provides a method for large block and intermediate block size support as well as a means for using common decoder architecture for different code rates and block size.
[0158] Another benefit of the present invention can be observed with respects to system
700 and system. In this system, the transmission of code blocks which are unsupported by an encoder, is made possible by the shortening, puncturing, and repeating of bits in the code blocks. This reduces the number of supported LDPC codes that need to be implemented.
[0159] A benefit of the present invention can be seen when looking at configuration
800. With these configurations, transmission of data has been optimized in regards to reliability and energy consumption. Optimization is achieved by choosing the transmitted block sizes such that the minimum transmitted size is maximized.
[0160] The final benefit of the present invention can be seen with respects to system
900. Using a common LDPC code to decode data of various lengths can be done while using a common piece of architecture. In previous methods, architecture was needed for each separate LDPC code, which added unwanted complexity to the hardware design.
[0161] The foregoing description of various preferred embodiments of the invention have been presented for purposes of illustration and description. It is not intended to be exhaustive or to limit the invention to the precise forms disclosed, and obviously many modifications and variations are possible in light of the above teaching. The example embodiments, as described above, were chosen and described in order to best explain the principles of the invention and its practical application to thereby enable others skilled in the art to best utilize the invention in various embodiments and with various modifications as are suited to the particular use contemplated. It is intended that the scope of the invention be defined by the claims appended hereto.