AN ADAPTIVE CODING SCHEME FOR OFDM WLANS WITH A
PRIORI CHANNEL STATE INFORMATION AT THE
TRANSMITTER
FIELD OF THE INVENTION
[01] The present invention relates to coded Orthogonal Frequency Division Multiplex (OFDM) wireless local area network (WLAN) systems. More particularly, the present invention relates to a coded OFDM WLAN system having channel state information (CSI) that is available at a transmitter.
BACKGROUND OF THE INVENTION
[02] Orthogonal Frequency Division Multiplex (OFDM) techniques, also referred to Discrete Multitone, have been applied in wireline communication applications (in which the communication is a wire, a fiber optic cable, and so forth), in which system performance is often channel-limited rather than noise-limited. With channel-limited situations for an OFDM system, optimum channel allocation of bits can be used for increasing error performance, data rate, or capacity when channel state information (CSI) is available at a transmitter. However, OFDM techniques utilized in wireline communication applications have not been extensively applied to wireless applications such as wireless local area networks (WLAN).
[03] Various approaches have been applied for improving system performance, a technique known as water-filling is used for improving error performance, data rate, or capacity. The water- filling technique is one approach in which the power level assigned to a sub-carrier (also known as a discrete tone) is dependent upon the frequency characteristics of the communications system. The "better" the frequency characteristics associated with the sub-carrier, the greater the power level that is assigned to the sub-carrier with a fixed power budget for the aggregate collection of sub-carriers. The bit loading technique is another approach for adaptively adjusting the number of bits allocated for each sub-
carrier. Multilevel coding is a technique in which the degree of protection for a bit position on each modulation symbol is dependent upon the error probability that is associated with the bit position.
[04] Nevertheless, what is needed is a way for optimizing channel allocation of bits in channel-limited system performance situations for an OFDM WLAN system when CSI is available at a transmitter, thereby improving error performance, data rate, or capacity by adjusting the channel code.
BRIEF SUMMARY OF THE INVENTION
[05] The present invention provides a way for optimizing channel allocation of bits in channel-limited system performance situations for an OFDM WLAN system when channel state information (CSI) is available at a transmitter, thereby improving error performance, data rate, or capacity by adjusting the channel code.
[06] The advantages ofthe present invention are provided by a method and a system for adaptively coding an orthogonal frequency division multiplexed (OFDM) signal at a transmitter in an OFDM WLAN system. According to the invention at least one block code, such as a Reed-Solomon code, is selected for handling errors based on a predetermined target data rate and a predetermined packet length for the OFDM signal. Next, a number of bits available in the OFDM signal for parity symbols that can be used for generating a selected integer number of OFDM symbols is determined. Preferably, the number of bits available in the OFDM signal for parity symbols are determined by subtracting a number of unprotected bits and coded bits from the determined number of bits required for transmitting the selected number of OFDM symbols. A number of erasure errors and a number of random errors over sub-carriers of the OFDM signal are then determined based on the determined number of parity symbols. Preferably, the number of erasure errors and a number of random errors over sub- carriers of the OFDM signal are determined based on a probability of error for bits in each symbol ofthe OFDM signal over the ordered sub-carriers. Lastly, the OFDM signal is encoded based on the determined number of erasure errors and
the determined number of random errors over the sub-carriers of the OFDM signal using each selected block code, such that the sub-carriers are ordered from a sub-carrier having a weakest channel amplitude to a channel having a strongest transmission amplitude based on channel state information that is available at the transmitter.
BRIEF DESCRIPTION OF THE DRAWINGS
[07] The present invention is illustrated by way of example and not limitation in the accompanying figures in which like reference numerals indicate similar elements and in which:
[08] Figure 1 shows a functional block diagram of an adaptive coding modulator according to the present invention for an OFDM WLAN transmitter in which CSI available at the transmitter;
[09] Figure 2 illustrates a flow diagram depicting a method for adaptively coding an orthogonal frequency division multiplexed (OFDM) signal at a transmitter in an OFDM system;
[10] Figure 3 shows a graph of Error Distribution at SNR = 20 dB for a channel having adaptive interleaving according to the present invention;
[11] Figure 4 shows a graph of Error Distribution at SNR = 20 dB for a channel without adaptive interleaving according to prior art;
[12] Figure 5 shows a graph of Error Distribution at SNR=25 dB for a channel having adaptive interleaving according to the present invention;
[13] Figure 6 shows a graph of Error Distribution at SNR = 25 dB for a channel without adaptive interleaving according to prior art;
[14] Figure 7 shows a graph of Bit Error Rate (BER) Performance for the 48Mbits/s mode of IEEE 802.1 la over a 24 tap Rayleigh Channel for an OFDM WLAN system incorporating the present invention; and
[15] Figure 8 shows a graph of Packet Error Rate (PER) Performance for the 48Mbits/s mode of IEEE 802.1 la over a 24 tap Rayleigh Channel for an OFDM WLAN system incorporating the present invention.
DETAILED DESCRIPTION OF THE INVENTION
[16] The present invention provides adaptive interleaving, or adaptive adjustment of channel coding parameters, in an OFDM WLAN system based on channel state information (CSI) that is available at a transmitter in the system. Channel coding parameters that are adaptively adjusted include the number of erasures or random errors, the number of parity symbols available for a packet (including stuff bit), the particular bits of a symbol that are to be protected, and partition of parity symbols among sub-carriers. In that regard, the present invention provides an adaptive interleaving approach that clusters errors together, and uses a block code, such as a Reed-Solomon (RS) code, for handling errors. The present invention is also applicable to other block codes, including a Bose, Chaudhuri, and Hocquenghem (BCH) code, a Fire code, a Golay code, a Hamming code, a Burton code, a Reed-Muller code, and a Goppa code. The approach of the present invention improves upon conventional "water-filling" techniques by not requiring multiple modulator/demodulators. Consequently, the present invention uses one RS encoder/decoder for generating different levels of protection for the sub-carriers.
[17] The adaptive interleaving technique of the present invention differs from conventional multilevel coding by using different levels of coding for each group of sub-carriers and corresponding symbols, while conventional multilevel coding assumes the channel conditions across all the sub-carriers are uniform. Thus, the present invention exploits the CSI at the transmitter, while multilevel does not use the CSI information.
[18] Figure 1 shows a functional block diagram of an adaptive coding modulator 100 according to the present invention for an OFDM WLAN transmitter in which CSI is available at the transmitter. Interleaving aims to distribute transmitted bits in time or frequency or both in order to achieve desirable bit error distribution after demodulation. Adaptive coding modulator 100 includes an interleaver 101, a modulator 102, a serial-to-parallel multiplexer 103, an Inverse Discrete Fourier Transformer (IDFT) 104 and a sub-carrier sorter 105. Data is input to interleaver 101. The output of interleaver 101 is modulated by modulator 102. In exemplary embodiment, the modulator utilizes 16-QAM modulation. The modulated output of modulator 102 is multiplexed by serial-to-parallel multiplexer 103 to form N modulated signals. The N modulated signals are converted to the time domain by IDFT 104 to form N time-domain signals. The N time-domain signals are sorted by sub-carrier sorter 105 based on CSI that is available at the transmitter. The sorted sub-carriers are ordered from weakest to strongest channel amplitudes and is discussed later in greater detail. The output of sub-carrier sorter 105 is then outputted for transmission as an OFDM signal in a well-known manner.
[19] According to the invention, the number of parity bits that are available for transmission can be determined from a given target data rate and a given packet length. Figure 2 illustrates a flow diagram depicting a method for adaptively coding an orthogonal frequency division multiplexed (OFDM) signal at a transmitter in an OFDM system according to the present invention.
[20] For example, in accordance with IEEE 802.11 a Standards, consider a target data rate of 48 Mbps data having 100 byte packets and a 64-QAM modulation, which contains 6 bits per symbol, and which requires a coding rate of 2/3. (As can be appreciated by one skilled in the art, the number of bits that are associated with a symbol can be different than 6 bits. In fact, the number of associated bits per symbol can vary from one bit to any number of bits per symbol.) If two RS(63,31 ,32) codes are chosen (RS denotes a Reed-Solomon code), a total of 64 parity symbols (each RS block supporting 32 parity symbols) are added that
are each 6 bits in length. The selection of block coding corresponds to step 201 in Figure 2. Thus, the code rate r for the selected RS code is:
8 * 100 800 r - ■ = 0.6757 .
8 * 100 + 32 * 2 * 6 1184
[21] The numerator for the ratio for code rate r represents the number of information bits in the packet, while the denominator represents the total number of bits including coded bits in the packet. Additionally, ther,e are 800 - 31 * 2 * 6 = 428 unprotected bits.
[22] The next step in the adaptive coding is to consider the additional parity that can be used, rather than random stuff bits, for generating an integer number of OFDM symbols, which consists of 48 sub-carriers that each have 6 bit symbols. Thus, part of the present invention specifies utilization of the stuff bits for error protection. Previously, utilization of stuff bits for error protection has not been considered because the current standard uses a convolutional code. Thus, to use the stuff bits for error protection, more information bits would be needed than is available in the packet. Even if such bits were available, the error rate would not improve. Hence, the bit error rate (BER) would remain the same, but the packet length would increase. This would lead to degradation in the packet error rate (PER) performance because the probability of a bit in a packet would increase in accordance to the percentage of increase in packet length. In the present case, RS codes (in general, block codes) easily lend themselves to increase the number of parity symbols per RS codeword.
[23] Now, the additional available bits for error protection N . is found as
1184
N pari •ty * 48 * 6 - 1184 = 256 tøte , (1)
48 * 6
where the symbol ' x~|" denotes the smallest integer greater than x. The
determination of additional available bits for error protection (in addition to
the 64 parity symbols associated with two RS code blocks) corresponds to step 203 shown in Figure 2. The additional available bits for error protection
require an additional RS code block as exemplified in the present example.
[24] More specifically, Equation (1) determines the number of bits required for transmitting an integer number of OFDM symbols and subtracts the number of unprotected bits and coded bits to find the additional available bits from error protection N parity . (Each OFDM symbol in this example corresponds to 1 symbol assigned to each ofthe 48 sub-carriers. With the example, the term "symbol" by itself should be interpreted as denoting a 6-bit symbol.) For this illustrative example, there are a little over 42 additional parity symbols that are available for error protection. This corresponds to a frame efficiency of 82.22%. In order to increase the frame efficiency, the number of symbols conveying information can be increased at the expense of fewer available parity symbols, thus reducing the corresponding error protection. In this example, there are a total of 32 * 2 + 42 = 106 parity symbols to protect five OFDM symbols.
[25] Another unique property of RS codes and block codes is the flexibility to specify an erasure location. An erasure is the location of a known error. In general, the error protection capability (in symbols) available is
[26] where N is the block length ofthe RS code, .AT is the length ofthe information block, Νerasures is the number of erasures supported by the Reed-Solomon code, and ΝRandomErrors is the number of random errors that can be supported by the Reed-Solomon code. The determination of Eprotection corresponds to step 205 in Figure 2. (With an erasure, the location ofthe error is determined while with a random error, the location of the error cannot be determined.) The adaptive algorithm ofthe present invention determines the number of erasures and random errors over the particular sub-carriers ofthe system. As an illustration, consider the error distribution shown in Figure 3 for 48 Mbps data rate at SΝR = 20 dB.
The dashed vertical lines in Figure 3 are decision boundaries that define decision regions for erasures (Region 21), random errors (Region 22) and unprotected bits (Region 23). Before decision regions can be defined, the channel response for the entire packet is sorted from weakest to strongest signal strength. The channel state information (CSI) is known at the transmitter. With the exemplary embodiment, the CSI is determined by sending a known sequence to the receiver. The receiver responds by sending channel transmission information to the transmitter for each sub-carrier. (It is assumed that the channel characteristics are sufficiently time stationary with respect to the time duration of data transmission.) The decision boundaries are determined by considering the bound on the error probability, i.e.,
P(x → x) ≤ e4N° (3)
where
[27] where pk is the known CSI, d2 is the square ofthe Hamming distance between x and x , Es is the energy of a symbol, and N0 is the noise level. Consequently, loose bounds can be found for the probability of error for the bits in each symbol over the sorted sub-carriers. The error probability is used for specifying the erasures on a first group ofthe weakest sub-carriers, random errors over a second group of moderately strong sub-carriers, and no protection for some uncodedbits over a third group of strongest sub-carriers. The boundaries separating Region 21, Region 22, and Region 23 (as shown in Figure 3) are chosen to transmit information bits (as conveyed in symbols) having a desired probability of bit error with the error protection ofthe selected block coding.
[28] In contrast to Figure 3, Figure 4 is a graph showing the error distribution for 48 Mbps data rate at SNR = 20 dB for a system that does not utilize the adaptive
interleaving of the present invention. Figure 5 is a graph showing the error distribution for 48 Mbps data rate at SNR = 25 dB. The dashed vertical lines in Figure 5 are decision boundaries that define decision regions for erasures (Region 41), random errors (Region 42) and unprotected bits (Region 43). In contrast to Figure 5, Figure 6 is a graph showing the error distribution for 48 Mbps data rate at SNR = 25 dB for a system that does not utilize the adaptive interleaving ofthe present invention.
[29] In both Figures 3 and 5, the first 52 assignments to sub-carriers are specified for erasures, the next 94 assignment to sub-carriers are specified for random errors for the 4 LSBs of a symbol, and the remaining sub-carriers are filled with remaining symbols from the RS block and uncoded symbols. The basic approach is to provide more robust coding (erasures) for the "weakest" sub-carriers and less robust coding (random errors) for the "stronger" sub-carriers. The sub-carriers are partitioned into groups in order to achieve the desired error rate for the associated frequency bandwidth spanning the collection of sub-carriers and for the block coding. Encoding the OFDM signal corresponds to step 207 in Figure 2. Thus, the adaptive coding block looks like
[RS1(1 :26) RS2(1 :26) RS3(1:94) RS2(27:63) RS1(27:63) Uncoded]
where the left bracket "[" corresponds with the weakest sub-carrier and the right bracket "]" corresponds to the strongest. The notation "RSx(y:z)" denotes the Xth RS block and the symbol assignment from the yth symbol ofthe RS block to the zth symbol ofthe RS block. The set of sub-carriers is partitioned into subsets, each subset corresponding to different degrees of received signal strength, e.g. "weakest" and "strongest". (In other words, the frequency characteristics over the frequency spectrum may not be uniform.) In the exemplary example, error coding for erasures is mapped to the "weakest" group of sub-carriers, error coding for random errors is mapped to the "next strongest" group of sub-carriers, and the uncoded symbols are mapped to the "strongest" group of sub-carriers.
[30] Figures 7 and 8 respectively show the bit error rate (BER) and packet error rate (PER) performances for the 48 Mbps data rate, 5 tap Rayleigh fading, with 100 byte packets using the coded bits ofthe present invention. In Figures 6 and 7, the performance for trellis-coded modulation using a rate 2/3 convolutional encoder (61 and 71 ) and performance (62 and 72) for block coded modulation and adaptive coding according to the present invention are shown. The major factor that determines the performance of the codes used by the present invention is coset selection based on detection ofthe coded bits. Hence, when the remaining bits are uncoded for both systems, then the coset selection will be the distinguishing factor in their relative performances. As shown in both Figures 7 and 8, the adaptive rate 2/3 coding system ofthe present invention is about 2 dB and 4 dB better than the convolutional code in BER and PER performance, respectively.
[31] While the invention has been described with respect to specific examples including presently preferred modes of carrying out the invention, those skilled in the art will appreciate that there are numerous variations and permutations ofthe above described systems and techniques that fall within the spirit and scope ofthe invention as set forth in the appended claims.