Reduced complexity coding system using iterative decoding

Abstract

A concatenated coding scheme, using an outer coder, interleaver, and the inner coder inherent in an FQPSK signal to form a coded FQPSK signal. The inner coder is modified to enable interative decoding of the outer code.

Description

SUMMARY

[0003] Properties of a channel affect the amount of data that can be handled by the channel. The so-called “Shannon limit” defines the theoretical limit of amount of data that a channel can carry.

[0004] Different techniques have been used to increase the data rate that can be handled by a channel. “Near Shannon Limit Error-Correcting Coding and Decoding: Turbo Codes,” by Berrou et al. ICC, pp 1064-1070, (1993), described a new “turbo code” technique that has revolutionized the field of error correcting codes.

[0005] Turbo codes have sufficient randomness to allow reliable communication over the channel at a high data rate near capacity. However, they still retain sufficient structure to allow practical encoding and decoding algorithms.

[0006] Feher's patented QPSK, or FQPSK, as described in detail in U.S. Pat. Nos. 4,567,602; 4,339,724; 4,644,565; 5,784,402; and 5,491,457 is a coded modulation scheme. The generic form of FQPSK is based on crosscorrelated phase-shift-keying. FQPSK maintains a nearly constant envelope, that is the maximum fluctuation in the envelope is around 0.18 dB. This is done by manipulating the pulse shapes of the in-phase “I” and quadrature “Q” signals using crosscorrelation mapping.

[0007] Many different variants of FQPSK are known, including FQPSK-B, which is a bandwidth limited form of FQPSK.

[0008] The price of this spectral efficiency of these coded modulation schemes may be a degradation in the bit error rate performance.

SUMMARY OF THE INVENTION

[0009] The present application teaches a new technique which allows additional power efficiency and bandwidth efficiency with a simple receiver architecture. This technique may use turbo coding techniques, along with a specially configured FQPSK encoder and/or decoder, to form a concatenated coded modulation scheme.

BRIEF DESCRIPTION OF THE DRAWINGS

[0010] These and other aspects will now be described in detail with reference to the accompanying drawings, wherein:

[0011] FIG. 1 shows a conceptual diagram of FQPSK;

[0012] FIG. 2 shows an alternative implementation of the baseband signals using mapping;

[0013] FIG. 3 shows full symbol waveforms of FQPSK;

[0014] FIG. 4 shows a trellis decoding interpretation used for a receiver;

[0015] FIG. 5 shows a bit error rate comparison of the different techniques;

[0016] FIG. 6 shows averaged waveforms for the simplified receiver;

[0017] FIG. 7 shows at trellis diagram for the simplified receiver;

[0018] FIG. 8 shows a simplified implementation of the baseband signals;

[0019] FIG. 9 shows a simplified receiver;

[0020] FIGS. 10 a -b show original and remapped encoders and trellises for the receiver to be used in concatenated schemes;

[0021] FIGS. 11-13 show embodiments of the improved transmitter and receiver; and

[0022] FIG. 14 shows a soft input soft output outer decoder for a rate ½ repetition outer code.

DETAILED DESCRIPTION

[0023] FQPSK in its standard form is similar to many phase-shift-keying techniques which had been previously used. A conceptual diagram of FQPSK is shown in FIG. 1. One advantage of the specific FQPSK system is a 3 db envelope reduction based on intentional but controlled crosscorrelation between the I and Q channels. This was described by half symbol mappings of the 16 possible combinations of I and Q channel waveforms that were present in the signal, into a specified set of 16 waveform combinations. These 16 waveform combinations were selected in a way that rendered the crosscorrelated output time continuous. The waveform combinations also had unit, normalized, envelopes at each of the I and Q uniform sampling instants.

[0024] Since the crosscorrelation mapping was based on half symbol characterization of the signal, there was no guarantee that the slope of the crosscorrelated output waveform would be continuous at the transitions between the half symbol points. In fact, a slope discontinuity may occur statistically one-quarter of the time.

[0025] In a copending patent application, it is suggested to structure the crosscorrelation mapping into a full symbol by symbol mapping, rather than a half symbol by half symbol representation. In fact, this technique also has the advantage of enabling data transitions on the I and Q channels to be described directly. Moreover, this enables a receiver for FQPSK which exploits the specific correlation (memory) that is introduced into the modulation scheme.

[0026] FIG. 2 illustrates the interpretation of FQPSK as a trellis-coded modulation. The input streams of data bits are split into time-aligned I and Q symbol streams at half the usual bit rate, so 1/Ts=½Tb. Each of these symbol streams is passed through specific rate ⅓ encoders. The rate ⅓ encoder 200 for the I stream 200 is different than the rate ⅓ encoder 240 for the Q stream 250. The output bits of these encoders are then considered to be grouped into one of three different categories.

[0027] The three categories include a first category of bits that only influence the choice of the signal in the same channel. A second category of bits only influence the choice of the signal in the other channel. A third category of bits influence choices of signals in both channels, that is represent crosscorrelation mapping.

[0028] Out of the bit sequences from the I encoder 200, the value I3 signal 201 is used to determine the signal that is transmitted on the I channel. The value Q0 signal 202, is used to determine the signal transmitted on the Q channel. The value I2, which is the same as Q1, signal 203, is used to determine both the signals transmitted on I and Q channels.

[0029] If d1n and dQn respectively denote the +1 and −1 I and Q data symbols in the nth transmission interval and 1$D_{ln}\stackrel{\Delta }{=}\left(1-d_{ln}\right)/2\mathrm{and}{D}_{\mathrm{Qn}}\stackrel{\Delta }{=}\left(1-d_{\mathrm{Qn}}\right)/2$

[0030] their (0,1) equivalents, then the mappings appropriate to the I and Q encoders of FIG. 2 are

I 0 =D Qn ⊕D Q,n−1 , Q 0 D l,n+1 ⊕D ln

I 1 =D Q,n−1 ⊕D Q,n−2 , Q 1 =D ln ⊕D I,n−1 =I 2

I 2 =D ln ⊕D I,n−1 , Q 2 =D Qn ⊕D Q,n−1 =I 0

I 3 =D ln , Q 3 =D Qn   (1)

[0031] These values correspond to the four, I channel coded bits which include two from the I encoder output and two from the Q encoder output. Analogously, it includes four, Q channel encoded bits.

[0032] The values i,j are used as binary coded decimal indices defined as follows:

i=I 3 ×23+I2×22+I1×21×I0×20

j=Q 3 ×23+Q2×22+Q1×21+Q0×20  (2)

[0033] The indices i and j may range between zero and 15. A set of baseband signals are shown in FIG. 3. These symbols may be defined as 2$\begin{array}{cc}\begin{array}{c}{s}_1\left(t\right)=\{\begin{array}{cc}A,& -T_s/2\le t\le 0\\ 1-\left(1-A\right){\cos }^2\frac{\pi t}{T_s},& 0\le t\le T_s/2\end{array},s_9\left(t\right)=-s_1\left(t\right)\\ s_2\left(t\right)=\{\begin{array}{cc}1-\left(1-A\right){\cos }^2\frac{\pi t}{T_s},& -T_s/2\le t\le 0\\ A,& 0\le t\le T_s/2\end{array},s_{10}\left(t\right)=-s_2\left(t\right)\\ s_3\left(t\right)=1-\left(1-A\right){\cos }^2\frac{\pi t}{T_s},-T_s/2\le t\le T_s/2s_{11}\left(t\right)=-s_3\left(t\right)\\ s_4\left(t\right)=A\sin \frac{\pi t}{T_s},-T_s/2\le t\le T_s/2,s_{12}\left(t\right)=-s_4\left(t\right)\\ s_5\left(t\right)=\{\begin{array}{cc}A\sin \frac{\pi t}{T_s},& -T_s/2\le t\le 0\\ \sin \frac{\pi t}{T_s},& 0\le t\le T_s/2\end{array},s_{13}\left(t\right)=-s_5\left(t\right)\\ s_6\left(t\right)=\{\begin{array}{cc}\sin \frac{\pi t}{T_s},& -T_s/2\le t\le 0\\ A\sin \frac{\pi t}{T_s},& 0\le t\le T_s/2\end{array},s_{14}\left(t\right)=-s_6\left(t\right)\\ s_7\left(t\right)=\sin \frac{\pi t}{T_s},-T_s/2\le t\le T_s/2,s_{15}\left(t\right)=-s_7\left(t\right)\end{array}& \left(3\right)\end{array}$

[0034] The pair of indices are used to select which of these baseband signals si(t), sj(t) will be transmitted respectively over the I and Q channels in any symbol interval.

[0035] For any value of A other than unity, certain waveforms will have a discontinuous slope at their midpoints (T=0). For example, it has been suggested that A should equal 1/sqrt(2) to produce minimum envelope fluctuation. When that happens, the waveforms 5 and 6 as well as their negatives 13 and 14, will have a discontinuous slope at those midpoints.

[0036] Finally, the I and Q baseband signals si(t)and sq(t)are offset by half a symbol relative to one another, and modulated onto the quadrature channels for transmission.

[0037] This trellis-coded characterization of FQPSK is, in principal, an M-ary signaling scheme. This means that a given pair of I and Q data symbols results in the transmission of a given pair of I and Q waveforms in each signaling interval. Restrictions are placed on the allowable sequences of waveforms that can be transmitted in either of these channels to produce continuous I and Q waveform sequences. The present inventors noticed that these restrictions on the transitional behavior of the transmitted signal results in the narrow spectrum characteristic of FQPSK. The inventors also noticed that the trellis coded structure of the transmitter suggests that an optimum receiver for FQPSK should be a form of trellis demodulator. It has been suggested to use of bank of 16 biased matched filters followed by a 16 state trellis demodulator. This configuration is shown in FIG. 4. The simulated bit error rate performance of this receiver is shown in FIG. 5 and compared with a conventional receiver as well as the performance of conventional uncoded QPSK.

[0038] This receiver may be relatively complex, and hence simplified configurations may be desirable. An averaged matched filter that is matched to the average of the 16 waveforms may replace the bank of 16 matched filters. A reduced complexity of the Viterbi receiver recognizes similarities in shape properties of certain members of the waveforms, and separates them into different groups. The waveforms s0-s3 are grouped as a first, composite waveform, with each four waveforms being similarly grouped as follows: 3$\begin{array}{cc}\begin{array}{cc}{q}_0\left(t\right)=\sum _{i=0}^3{s}_i\left(t\right),& q_1\left(t\right)=\sum _{i=4}^7{s}_i\left(t\right),\\ q_2\left(t\right)=\sum _{i=8}^{11}{s}_i\left(t\right)=-q_0\left(t\right),& q_3\left(t\right)=\sum _{i=12}^{15}{s}_i\left(t\right)=-q_1\left(t\right)\end{array}& \left(4\right)\end{array}$

[0039] The waveform assignments of the group members are then replaced by their corresponding average waveform that is, any of s0 to s3 become q0 to q3. This causes the crosscorrelation between the I and Q channels to effectively disappear. Effectively, the I channel signal is selected based on only the I encoder output bits, and the Q channel signal is based on only the Q encoder output bits. When this happens, then the trellis coded structure decouples into two independent I and Q two state trellises; see FIG. 7. The transmitter simplifies into the FIG. 8 structure with a corresponding optimum receiver being shown in FIG. 9. The I and Q decisions are no longer produced jointly, but rather are produced separately by individual Viterbi techniques acting on energy based correlations from the I and Q modulated signals. The degradation in bit error rate relative to the optimum receiver may be compensated by the significant reduction in complexity of the receiver.

[0040] FQPSK, as described above is a convolutional coded modulation. It is recognized by the inventors that a potentially large coding gain may be achievable using iterative/recursive encoding and decoding of concatenated codes with a soft input soft output a posteriori probability algorithm.

[0041] The techniques of concatenated codes are well-known. In general, this system has two encoders: an outer coder and an inner coder separated by an interleaver. A serial concatenated code operates serially, while a parallel concatenated code operates in parallel. An outer encoder receives the uncoded data. The outer coder can be an (n,k) binary linear encoder where n>k. The means that the encoder 200 accepts as input a block u of k data bits. It produces an output block v of n data bits. In its simplest form, the outer coder may be a repetition coder. The outer coder codes data with a rate that is less than 1, and may be, for example, ½ or ⅓.

[0042] The interleaver 220 performs a fixed pseudo-random permutation of the block v, yielding a block w having the same length as v. The permutation can be an identity matrix, where the output becomes identically the same as the input. Alternately and more preferably, the permutation rearranges the bits in a specified way.

[0043] The inner encoder 210 is a linear rate 1 encoder.

[0044] According to a present system, this technique is applied to FQPSK. It is recognized that the inherent coding that is carried out in FQPSK may supply the inner code for the iterative concatenated code. In the embodiments, the outer coded signal is applied to a FQPSK system which may use the simplified receiver of FIG. 9. The two state Viterbi algorithms are replaced with two state soft input-soft output (SISO) Max-log algorithms as described in the literature. These may be considered as modified soft output Viterbi algorithms. An interleaving/deinterleaving process is applied between the inner and outer codes. A coding gain from this interleaving process can be obtained by remapping the I and Q FQPSK inner codes from nonrecursive into recursive type codes, using other known techniques. Figures 10a-10b show the original I and Q encoders and the remapped I and Q encoders for this purpose.

[0045] This remapping provides recursiveness for the parts of the FQPSK encoders that are matched to the reduced two state soft input-soft output decoder for the inner code.

[0046] The remapped encoders would produce different baseband waveforms. However, the allowable FQPSK encoder output sequences would remain the same. Therefore, both the envelope and spectral characteristics of the modulated signal would be identical to those produced by the FQPSK signal in the transmitter in FIG. 2.

[0047] When an outer code is added, an interleaver is used which has a size that is large enough to approximately output an uncorrelated sequence.

[0048] FIGS. 11-13 shows three different embodiments of applying an outer code to the FQPSK modulator/demodulator using a concatenated system with iterative decoding. A number N of input bits 1100 are applied to a demultiplexer 1105 that divides the bits between a pair of outer encoders 1110, 1115 of rate R. These effectively form the I and Q channel bit streams. Each of the outer-encoded bits are applied to interleavers 1120, 1121. The I and Q channels are then applied to an FQPSK inner code modulator 1125 which forms the inner code of such a system. The thus coded stream 1130 is transmitted over the channel 1135. The receiver includes a matched filter bank 1140 with biases, that produces I and Q output channels. Each of the channels goes through a soft input-soft output FQPSK demodulator 1145, 1146 whose output is coupled to a deinterleaver 1150,1151. The resulting I and Q baseband signals are multiplexed in 1165 to provide the decoded output bits 1170.

[0049] In this and the other similar embodiments, the energy biased matched filter bank 1140 provides for branch metrics per I and Q channel for the simplified, two state soft in soft out FQPSK coders. The decoders 1145,1146 provide extrinsics associated with the FQPSK encoder input bits to the outer coder via the deinterleavers 1150, 1151. These are applied to the outer decoders 1155,1156 to provide new versions of the received extrinsics by using the code constraint as an output extrinsic through the interleavers. The other outputs are fed back to the inputs of the demodulators 1145,1146.

[0050] In operation, the process may repeat/iterate several times. At the end of the final iteration, the output of the outer decoders 1155, 1156 are hard limited in order to produce decisions on the bits.

[0051] An alternative system shown in FIG. 12 receives the input bits 1100 directly to an outer encoder 1205, and interleaver 1210 whose output is demultiplexed by 1215 and applied to the FQPSK inner code modulator 1220. This system may use an analogous receiver with the matched filter bank 1140, applied to similar demodulators 1145, 1146. The demodulated outputs are multiplexed by 1250, and deinterleaved 1255, and then outer decoded 1260. The feedback loop in this system uses an interleaver 1265 and demultiplexer 1270 to provide the I and Q channels.

[0052] FIG. 13 shows a parallel concatenated coding system, in which a rate 1 outer encoder is formed from the input bitstream and the interleaved input bitstream as in a turbo code. The input bits 1300 are split with one set of bits being interleaved by 1305. The input bits and interleaved input bits are applied in parallel to the inner coder/FQPSK module which outputs a coded bitstream 1315 that is applied to the channel 1320. Data from the channel is received into a matched filter bank type receiver 135 that iteratively calculates the output.

[0053] The outer coder may simply be a rate ½ repetition outer coder, with a block interleaver of size n. The outer decoder 1155 may be significantly simplified for the repetition code. For this code, the outer decoder may simply swap the order of successive pairs of bits as shown in FIG. 14.

[0054] Computer simulations of this system show an improvement of 3.75 dB at up bit error rate of 10−5.

[0055] Although only a few embodiments have been disclosed in detail above, other modifications are possible. All such modifications are intended to being comps within the following claims, in which:

Claims

1. A method of coding, comprising: passing a sequence of bits to be coded through an outer encoder to produce outer encoded bits; and forming a modified FQPSK modulation on said outer encoded bits of the type which applies an inner code to said outer coded bits.

2. A method as in claim 1, wherein said forming comprises remapping the I and Q FQPSK inner codes from nonrecursive into recursive type codes.

3. A method, comprising: outer coding a plurality of bits to form a repetition code; interleaving said bits according to a specified interleaving scheme; and remapped FQPSK modulating the repetition coded bits to form a concatenated code that will be transmitted over a channel.

4. A method as in claim 3, wherein said remapped concatenated code is a parallel concatenated code.

5. A method as in claim 3, wherein said concatenated code is a serial concatenated code.

6. A method as in claim 3, further comprising decoding said concatenated code using an FQPSK demodulator in an iterative loop.

7. A method as in claim 3, wherein said outer coding comprises the multiplexing input bits into in phase and quadrature channels, and separately outer encoding said in phase and quadrature channels.

8. A method as in claim 3, wherein said outer encoding comprises outer encoding input bits and then demultiplexing said input bits into in phase and quadrature channels.

9. A receiver, comprising: a modified FQPSK demodulator, an iterative decoder, which iteratively demodulates and decodes an input signal.

10. A receiver as in claim 9 further comprising a deinterleaver.

11. A receiver as in claim 9, wherein said receiver includes a matched filter bank.

12. A method, comprising: using a modified FQPSK coding system to code according to an iterative coding scheme.

13. A method as in claim 12, wherein said iterative coding scheme is a turbo coding type scheme.

14. A method as in claim 12, wherein said to using comprises outer coding a signal, and interleaving the signal.

15. A method as in claim 13, wherein said turbo coding scheme is a serial type concatenated coding scheme.

16. A method as in claim 13, wherein said turbo coding scheme is a parallel type concatenated coding scheme.

17. A method, comprising decoding a modified FQPSK coded signal according to an iterative decoding scheme.

18. A method as in claim 17, wherein said decoding uses a matched filter bank.

19. A method as in claim 17, wherein said decoding is a turbo decoding scheme using a deinterleaver and an outer decoder.

20. A method as in claim 17, wherein an FQPSK demodulation also includes inner decoding.

21. A method as in claim 17, where the outer code can be of the repetition, convolutional, or block type.