DIGITAL AUDIO DIFFUSION METHOD USING PERFORMATIVE CONVOLUTIONAL CODE
DESCRIPTION OF THE INVENTION This invention relates to diffusion, and more particularly, correction to error in digital audio broadcasting (DAD) in channel band (IBO) and broadcast systems that use such error correction. Digital audio broadcasting is a means to provide digital quality audio, superior to existing analog broadcast formats. Both DAD ECEB of AM and FM can be transmitted in a hybrid format where the digitally modulated signal coexists with the analog signal currently in broadcast. ECEB requires non-new spectral locations since each DAD signal is simultaneously transmitted within the same spectral mask of an existing channel location. ECEB promotes spectrum economics while allowing broadcasters to provide digital quality audio to their present listener base. FM ECEB systems using a hybrid modulation format have been the subject of several United States patents including patents no. 5,465,396; 5,315,583; 5,278,844 and 5,278,826. In addition a commonly assigned United States Patent NO. 5,956,624 describes an FM ECEB DAD system. An orthogonal frequency division multiplexing (OFDM) technique for DAD ECEB has been described. The OFDM signals consist of orthogonally separated carriers all modulated in a common symbol ratio. The frequency spacing for rectangular pulse symbols (for example, BPSK, QPSK, 8PSK or QAM) is equal to the symbol ratio. For the ECEB transmission of FM / DAD signals, a group of OFDM subcarriers is placed within approximately 100 kHz at 200 kHz on either side of a coexisting analog FM carrier. The power of DAD (upper or lower sideband) is set to approximately -25 dB relative to the FM signal. The level and spectral occupancy of the DAD signal is set to limit the interference to its FM host while providing a signal-to-noise (SNR) relationship for the DAD subcarriers. The first signals adjacent to + -200 kHz from the FM carrier can corrupt the DAD signal. However, at any particular location within a station coating area, it is unlikely that both adjacent firsts will significantly interfere with DAD. Therefore the upper and lower DAD sidebands carry the same redundant information in such a way that only one of the sideband is necessary to communicate the information. The inherent advantages of OFDM include robustness in the presence of multipath interference, and tolerance to non-Gaussian noise or short-term slots due to selective fading. Fault correction (FEC) and interleaving improve the reliability of digital information transmitted over a corrupted channel. The correction to error using punched codes has been suggested by others. For example, the United States Patent NO. 5,197,061 suggests using drilling techniques for different levels of error protection. Also see S. Kallel, "Complementary Punctured Convolution (CPC) Codes and Their Applications", IEEE Trans. Comm. , Vol. 43, No. 6, p. 2005-2009, June 1995. Techniques of complementary drilling convolution Fec (CPC) codes are developed by automatic repetition request (ARQ) schemes where retransmissions are encoded using complementary codes instead of simply retransmitting the same coded sequence. The CPC codes can be constructed according to previously published perforation techniques, for example Y. Yasuda, K. Kashiki, Y. Hirata, "High Rate Punctured Convolutional Codes for Soft Decision Viterbi Decoding", IEEE Trans. Comm. Vol. 32, # 3, Mar. 1984; and J. Hagenauer. "Rate-Compatible Punctured Convolutional Codes (RCPC Codes) and Their Applications," IEEE Trans. Comm. Vol. 36, No. 4, p. 389-400, April, 1988. It is known that periodic bit drilling from a convolutional code using Viterbi decoding is an effective means of creating higher proportion convolutional codes. Perforated convolutional codes compatible with the ratio (RCPC) as a mechanism for adjusting coding gain and bit energy as a function of channel capacity in a practical efficient manner, see the reference by former Hagenauer or M. Kim, "On Systematic Punctured Convolutional Codes ", IEEE Trans. Comm., Vol. 45, No. 2, p. 133-139, Feb, 1997. This is useful in a point-to-point (non-broadcast) automatic repeat request (ARQ) system where the proposed receiver evaluates its ratio of signal to noise power (Eb / No) and communicates its I wish the transmitter (via a return path) to increase or decrease the power per bit (Eb) and gain coding. The transmitter responds by setting 1 its coding ratio R. This is done with a perforated convolutional code where the transmission of all the bits typically employs a ratio code K = 7, R = 1/2"industrial standard", for example. It is assumed in this non-perforated case that the coding gain and maximum Eb is achieved. To improve the spectral and / or power efficiency, the transmitter may choose to eliminate (drill on the receiver request, for example) the transmission of some of the encoded bits, resulting in a higher rate code. This perforation has the effect of decreasing the coding gain and effective Eb relative to the original non-perforated code; however, this drilled code may still be sufficient to successfully communicate information about the channel in a more efficient manner. For better behavior at a given code rate, a particular bit pattern is drilled in the encoded sequence. Unfortunately, the drilling pattern for higher proportion codes does not include all the punched bits for lower ratio codes. Haganauer shows that drilling patterns for their RCPC codes can include all drilling for lower ratio codes with less loss compared to optimal drilling patterns, but incompatible in proportion. Therefore the code rate can be increased from the original R = l / 2 code by simply drilling more of the punctable bits of the same pattern. The highest ratio codes are a subset of the bits of the lower ratio codes. The interference environment in ECEB DAD channels of FM VHF band is generally such that a DAD channel can be dichotomized in the following two sub-groups of sub-channels: (a) a reliable part composed of regions of relatively interference-free spectrum from other signals of station, characterized as being a thermal noise or limited antecedent, with multipath fading as a damage; and (b) an unreliable part composed of spectrum regions with intermittent intervals of heavy interference which corrupts the transmitted bits during those intervals, but it is at other times (or for most geographic locations) similar to the reliable part described previously. ECEB DAD AM band can be similarly characterized. The prior art uses one of the two fundamental strategies for transmitting data in this environment: (1) simply not using the unreliable part of the channel, in this way those times during which the unreliable part is clear and usable are essentially discarded; or (2) use sufficiently low proportion code (and proportion of appropriately increased encoded bits) to guarantee the required bit error ratio (SEE), and distribute the increased bandwidth between both reliable and unreliable portions of the spectrum evenly. This is done by uniformly allocating bits to OFDM carriers in an OFDM system, or increasing the proportion of prime bits of an individual bearer system. This uses the unreliable part of the channel, but also incurs a BER penalty (possibly catastrophic) when severe interference occurs in the unreliable part of the channel. Depending on the interference, the second alternative may or may not be better first. This invention handles non-uniform interference through special coding and error handling to achieve better robust behavior. The diffusion method of the invention encodes program material using convolutional codes having non-drillable bits and drillable bits and modulates multiplexed frequency division bearer signals orthogonally with convolution codes. The non-pierceable bits are carried by a first group of carriers and the pierceable bits are carried by a second group of carriers, where the first group of carrier signals is less susceptible to interference then the second group of barrier signals. The bearer signals are then broadcast to receivers which determine whether the bearers in the second group have been corrupted and eradicate drivable bits carried by any of the bearers that have been determined to be corrupted. These punched codes are subsequently decoded to retrieve the program material. The invention also encompasses transmitters and receivers that operate in accordance with the method of the invention. This invention provides a technique FEC coding which results in an improved bit error ratio in the interference environment of a digital band-in-channel audio broadcasting system using orthogonal frequency division multiplexing. This mitigates the effects of interference from non-uniform interference in orthogonal frequency division multiplexing broadcast systems. BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a schematic representation of the frequency spectrum of a guest FM radio station that broadcasts a DAD signal, with a first adjacent channel interference device; Figure 2 is a schematic representation of a convolution coder K = 7, R = l / 2; Figure 3 is a simplified functional block diagram of a transmitter and receiver which operates in accordance with the method of the invention; Figure 4 is a functional block diagram showing the mapping and processing of bits through a receiver, deinterver, and error correction decoder; < Figure 5 is a schematic representation of a convolution coder K = 7, R = l / 3; and Figures 6 and 6a are schematic representations of an array of block codes that can be used in the invention. The particular request of the preferred embodiment is for a band-channel digital audio broadcasting (DAD) system (ECEB) where the external OFDM subcarriers in addition to the host FM bearer are sometimes subjected to destructive interference from the first adjacent and, possibly, the second adjacent channels. An illustration of the interference environment is shown (not to scale) in a typical FM band DAD scenario in Figure 1. Figure 1 is a schematic representation of the frequency locations (spectral placement) and relative power spectral density of the signal components for a DCE ECEB hybrid FM signal 10 which may use the present invention. The hybrid format includes the conventional FM stereo 12 analog signal having a spectral power density represented by the triangular conformation 14 generally placed in a band? central, or center frequency, portion of an FM 18 band channel. The power spectral density (PSD) of a typical analog FM broadcast signal is almost triangular with a slope of approximately -0.35 dB / kHz from the frequency central. A plurality of uniformly separated digitally modulated subcarriers are placed on either side of the analog FM signal, on an upper side band 20 and a lower side band 22, and are transmitted concurrently with the analog FM signal. The signals from an adjacent FM channel 24 (ie, the first adjacent FM signals), if present, can be centered at a spacing of 200 kHz from the center of the channel of interest. The first adjacent FM band station bearers spectrally overlap the DAD channel, on the average, as their respective FM bearers sweep in and out of the DAD channel. This potential spectral overlap can result in corrupting the orthogonal frequency division multiplexed bearers that are located in the portion of the spectrum subject to overlap. It is therefore apparent that the carriers resting near the ends of the upper sideband (in regions 26 and 28) are more susceptible to interference than those carriers resting near the center of the upper sideband (in region 30). The lower sideband may be subjected to similar spectrum that overlaps from the analog FM signal and the first adjacent FM signal on the other side of the channel of interest. Even carriers that rest within regions interfered with regions can be made to improve the total channel. In the hybrid system the total DAD power in the OFDM modulated subcarriers in each sideband is set at approximately -25 dB in relation to its host analogue FM power. The method of this invention employs a perforated code and segregates the drillable bits from the non-drillable bits by placing the bits punched in the unreliable part of the channel. Perforable codes are codes that use bits, called drillable bits that, if erased, still allow recovery of the encoded data but with reduced error correctness compared to the original non-perforated code. The other bits, those that can not be deleted without substantial loss of coding gain, are called non-drillable bits. The drillable bits of the non-drillable bits are segregated by placing the drillable bits in the non-reliable part of the channel. Thus, when those bits get to be erased in the receiver, the behavior of the total channel is not so bad that it can contribute to reduce the VER of the reliable part of the channel, improving its previous behavior than the reliable part of the channel alone. In effect, this scheme allows dynamically clear periods of time of the unreliable part of the channel to be used to improve the BER behavior of the total channel1, without being penalized with increased BER during interfered periods. The technique described herein allows the receiver to adaptively delete the soft symbols from the particular OFDM carriers after the receiver has determined that the interference is sufficiently high to corrupt them. The particular FEC coding technique employed herein exploits the ability to drill or delete particular bits without excessive loss in coding gain. The RCPC coding technique of the previous reference paper by Hagenauer can be modified for application in an OFDM diffusion channel where the interference on the subchannels is not uniform, but predictable through non-uniform interference estimation (not white) or noise in any individual receiver. In this case, the transmitter broadcasts all the encoded (non-perforated) bits. However, the convolutionally coded bits are arranged in such a way that the possible punched bits will be transmitted in OFDM subcarriers that may be more vulnerable to corruption. The non-perforated encoded bits can be transmitted on the most reliable subcarriers. Any particular receiver can assess their individual interference situation, particularly on the most vulnerable subcarriers. If the receiver estimates high enough interference for these subcarriers, then the subcarrier may decide to clear the bits from these corrupted subcarriers before decoding. Erasing consists of setting the soft decision quantities of the deleted bits to zero. Since the erasures are made in the drilled bits, the same effect is achieved as drilled, but without involving the transmitter. Selective erasure in the corrupted bit receiver using a prior interference evaluation information can significantly improve the behavior. The OFDM transmission is unique in this aspect where knowledge through the estimation of non-uniform (non-white) interference or noise can be used to adaptively improve the FEC decoding behavior. Additional improvements or variations are possible in this concept. Some improvement in the behavior can be achieved through the appropriate "heavy" of the drilled bits instead of deleting at zero magnitude. This is possible in the present invention for the diffusion system compared to the previous non-diffusion systems where the bits are drilled in the transmitter. Ideally the appropriate weights in the soft bits arriving from each subcarrier should be in proportion to the signal to noise ratio (SNR) for each subcarrier (assuming additive white gaussian noise, AWGN). However, estimation errors in a non-Gaussian dynamic interference environment can reduce the potential effectiveness of this technique. The flexibility in the use of bits is increased since the "broadcaster" may wish to use the punched bits for some type of in-band signaling. In this case, some of the subcarriers that carry punched bits can be replaced with other data. This modification in format must also be broadcast to all receivers (for example, via a bit of mode control within the message format) in such a way that the receivers can punch these bits after decoding. Obviously this option can reduce the robustness of the resulting perforated data, but the broadcaster must consider this concession. A special type of punched code includes some systematic codes. Systematic codes include the sequence of input data as part of the output sequence, more additional parity bits. Clearly, with systematic codes, all parity bits can be erased and the encoded data can have a bit error ratio (BER) not so bad that not using codes at all. But when the parity bits are not all drilled, the total code has coding gain that results in BER behavior better than not using codes at all. Although it is well known that non-systematic codes perform systematic codes for low proportions (eg R = l / 2), it has been shown that systematic codes of high proportion seem to behave better than non-systematic perforated convolutional codes (see the document cited above). by Kim). Systematic codes are characterized by having an output that is a replica of the information sequence entry. A systematic feedback encoder can be constructed from a non-systematic feeder encoder with identical distance properties (ie, the same error correction behavior) of the unsystematic feed encoder where the feedback is implemented with a polynomial division operation binary A schematic representation of a convolutional ratio encoder is presented in Figure 2. The shift register 32 receives input data bits in line 34. These input data bits are received in a proportion B and are representative of the program to be transmitted. Such program material may include, for example, audio information representative of spoken or musical signals, and / or data. Touches are used to direct the information bits in the change recorder for module 2 aggregates 36 or 38. The selected touches are for illustration only. A seven-stage change recorder with appropriate touch points can be used for a code K = 7, R = l / 2. The output of added module 2 on line 40 contains punched encoded bits and the output of added module 2 on line 42 contains non-punched bits. A switch 44 recirculates in a proportion 2B for a proportion code ^ _. This produces the output bits coded on line 46 at a bit rate of 2B. Figure 3 is a simplified block diagram of a transmitter 48 and a receiver 50. The transmitter receives the program material, which may include, for example, stereo audio signals on lines 52 and 54 on line 60. Perforated code includes both non-drilled and drilled bits. A plurality of carriers are produced by the oscillator 62 and these carriers are then modulated by the code bits by the modulator 64 to produce a frequency division multiplexed signal orthogonally on the line 66 which is transmitted by the antenna form 68 to the receiver . The antenna 70 of the receiver receives the OFDM carriers. The demodulator 72 extracts the code from the OFDM carriers and the decoder 74 converts the code back to the program material that can be supplied to an output device such as a speaker or screen 78. Figure 4 is a further functional block diagram detailed showing the mapping and processing of bits through a portion of a receiver operating in accordance with the method of the invention. A plurality of OFDM carriers 80 are received and converted to bit streams on lines 82 by circuit 84 of the receiver. Circuit 84 includes a digitizer, bearer synchronization, symbol synchronization, and coupled filters all operating in accordance with well-known techniques to produce bit streams on line 82. Editor 86 detects bits and clears certain bits of punched bits (or reduce the weights of those bits) according to the interference level of the bearers used to transmit the bits to produce bit streams edited on lines 88. Block 90 shows that the bit streams are delocalized from the bearers and supplied to several intercalators 92, 94 and 96. Block 98 shows that a synchronous word is located in most trustworthy carriers. The outputs of the deinterleavers are multiplexed to a stream of individual bits as shown in block 100. A Viterbi decoder 102 decodes the individual bit stream. A calibrated delay is added in block 104 to allow mixing of the FM digital broadcast signal. The delayed signal is then passed to modern demarcation block 106 for further processing. The soft decision Viterbi decoding with optimal soft decision (near) heavy for maximum ratio combination (MRC) for differentially detected QPSK subcarrier symbols is used to minimize losses on the channel. A CPC code that can be used in this invention can be constructed at the start with a 1/3 industrial standard ratio convolutional code. A schematic representation of a 1/3 ratio convolutional encoder is presented in Figure 5. The change recorder 108 receives input data bits in line 110. These input data bits are received in a proportion B and are representative of the program material to be transmitted. Such program material may include, for example, audio information representative of spoken or musical signals, and / or data. Touches are used to direct the information bits in the recorder to module 2 aggregates 112, 114 or 116. The selected touches are for illustration only. A recorder gives seven-stage change with appropriate untwisting points can be used for a code K = 7, R = l / 3. The output of module 2 aggregates contains punched encoded bits and non-punched bits. A switch 118 recirculates at a ratio 3B for a 1/3 ratio code. This produces the output bits coded on line 120 at a bit rate of 3B. <; The 1/3 ratio convolution encoder of Figure 5 can be observed as producing 3 streams of encoded bits (Gl, G2 and G3), each of the same proportion as the input. The combination of these 3 bit streams produces the coded output sequence R = l / 3. To create a pair of complementary codes, for example, a subgroup of the exit code bits is assigned to the lower sideband DAD and a different (complementary) subgroup is assigned to the upper lateral band. Each subgroup must contain at least the same proportion of bits as the input proportion information, plus some additional bits to provide some coding gain. The encoded bit mask of a Drill Pattern array is shown as:
G1o G 1j sx- GI3 G2, ^ G2I s- »G3o Q3l G 'ü33
the drilling pattern matrix represents the encoding output symbols on each group of 4 information bits. Therefore the output symbols are identified and indexed in module 4. A logic 1 in any of the 12 locations of the mask indicates that that particular bit is used. Otherwise, a zero logic indicates that the bit is not used. This bit pattern can be chosen based on the best known R = 4/5 drilling pattern, or a RCPC code pattern. However, after a sideband is defined in this way, there is little flexibility in choosing 2
the bits for the opposite sideband since they must be chosen from the punched (complementary) bits to achieve the maximum coding gain when they are combined to form the base code. Fortunately, computer analysis and stimulation have verified that good complementary codes exist. For example, the bit patterns shown above produce very good behavior when Gl = 133, G2 = 171, and G3 = 165 where the generator connections are represented by standard octal notation. A pair of complementary drilling patterns can be shown, one for the upper DAD sideband and one for the lower DAD sideband as:
-0 1 1 0-, 1 0 0 1 PPL = | 1 0 0 1. PPU = ¡0 I i or i '• 0 0 1 0 1 o or 0 <
Individually these drill patterns define the pair of 4/5 ratio codes. The pair of complementary 4/5 ratio codes can be combined to form the 2/5 ratio base code as shown below. Each of the 4/5 ratio codes have a free distance of dr = 4 with error weight information Cd = 10. The combined ratio code 2/5 produces dr = ll with c_ = 8. Note that only half of the G3 bits are used in this CPC code. The drilling pattern for the original 2/5 code is:
I I 1 1 PPL-t-PPU = | 1i 1i 1i 1i \ 1 0 1 o
Optionally, the drilled bits of the 2/5 ratio code can be transmitted to produce a pair of CPC codes of 2/3 ratio with dr = 6, such as:
O l i o- i o or i; I 00 1 'PPUop = 0 1 1 0' or i i o: i o or i ••
Of course, the base code is the 1/3 unprocessed ratio code with dr = 14. A 4/5 ratio code in each side band requires 25% additional bits. A method for dividing bits to the sidebands can be represented as: Lower sideband Upper sideband G3, G20 G23 Gli Gl2 Glc Gl3 G2_ G22 G30 The above representation shows the relative spectral locations of the encoded bits. These spectral locations are maintained after interleaving by channeling the interleaver into different fractions that are mapped to the appropriate subcarriers in each sideband. The most expandable encoding bits are placed on the external OFDM subcarriers. The expandable bits contribute at least to the free distance or coding gain of the combined code. The optional G3 bits can be placed in the internal carriers closest to the host FM spectrum. The analysis and simulation have shown that this fractional interleaver performs random interleaving under typical interference scenarios. The use of CPC coding techniques as well as interleaving over time can also improve the behavior. A line interleaver arrangement 255 may be established per column 456 to maintain the bits produced by the convolutional encoder. A pictorial diagram of the layout of the interposer is presented in Figures 6 and 6a. Each line of the layout of the interleaver 122 maintains the code bits to be modulated in an OFDM symbol in parallel. Line 256 is reserved for the synchronous word of modem structure. Each pair of columns is assigned in the phase and quadrature modulation QPSK. Additional subcarriers may be used outside the intercalary for pilot transmission or other data applications. The encoding bits are written in the interleaver arrangement in a particular pattern. The layout is read, line by line, providing the data source for the OFDM symbols in parallel. The interleaver partition assignments can be displayed as:
7 3 4 g > torque assignment = 'i 9 10 2 i l l 6 0 5 / \ partitions =. { 0 1 2 3 4 5 FM 6 7 S 9 10 1 1)
The interleaver can be implemented by first assigning the coding bits (module 12 Index) of the drill pattern to the column partitions of the subcarrier 12. This is illustrated above using the partition index to identify the interleaver partitions corresponding to pattern bits of drilling. The ordering is in the range of 0 to 11 on the lower frequency subcarriers to higher frequency subcarriers to represent the partitions of the subcarrier 12. Each partition is comprised of 38 columns and carries proposed code bits for 19 subcarriers, where the real components and imaginary of a particular subcarrier are identified as separate adjacent columns. The total interleaver consisting of 12 partitions has 456 columns. The most external subcarriers are identified as columns 0.1 and 454.455. Columns 190 to 265 carry the optional perforated bits closest to the FM host spectrum. A portion of the interleaver layout can be displayed (lines 0 to 17, and columns 0 to 8) show the spacing of the partition index "K" as:
9120 9135 9150 9165 9180 9195 9210 9225 9240 I 16 31 46 61 76 91 106 121
Each partition is further divided into 15 blocks of 17 lines each. These blocks facilitate interleaving over time by separating code bits, which correspond to adjacent coded information bits, by the number of lines in a block. The line and column indices of the interleaver, line and column layout, respectively, are calculated using the following expressions:
co l = m od CO LS + 3 8 • part. ™ BL O CK $)
where the constants of the size of the interleaver are R0WS = 255, COLS = 38, BL0CKS = 15, and part is the partition
(part = 0, l, .... ll) of the drilling pattern K. A portion of the interleaver arrangement (Figure 6) shows that the consecutive values of the drilling pattern index k are separated at both time and frequency. This invention allows a receiver to mitigate the effects of interference from non-uniform interference in an orthogonal frequency division multiple diffusion system. The preferred embodiment of the invention relates to a channel-band digital audio broadcast (DAD) system (ECEB) where the external OFDM subcarriers in addition to the host FM bearer are sometimes subjected to destructive interference from the first adjacent and possibly, the second adjacent channels. The technique described herein allows the receiver to adaptively delete the soft symbols of the particular OFDH carriers after it has been determined that the interference is sufficiently high to corrupt them. The particular FEC coding technique employed herein exploits the ability to drill or erase particular bits without excessive loss in coding gain. DAD ECEB is an ideal candidate for the application of CPC codes since the digital DAD transmission is carried out on two lateral bands (upper side band and lower side band) that are potentially damaged by almost independent interference devices with independent fading. If a sideband is completely corrupted by a strong first adjacent FM signal in the vicinity of the receiver, the sideband should be independently decodable by the receiver. Therefore, each sideband must be coded with an independently decodable FEC code. However, when both sidebands contain useful information that is not completely corrupted by an interference device, then the CPC codes provide additional coding gain earlier than that achieved by combining the two sides. Additionally, the OFDM interleaving techniques have been developed to perform single interference and selective fading characteristics of the FM ECEB DAD channel. This invention exploits the interleaving in time to mitigate the effects of flat fades (or broadband) over times of multiple symbols, and exploits a priori knowledge of non-uniform sub-channel interference statistics. The latter has resulted in the careful placement of the code bits on the subcarriers, and the selection of the CPC codes for the hybrid ECEB DAD FM application. While the present invention has been described in terms of what is believed to be its preferred embodiments, it will be apparent to one skilled in the art "that various changes may be made to the embodiments described above without departing from the scope of the invention as set forth in The following claims.