412
`
`IEEE TRANSACTIONS ON BROADCASTING, VOL 43, NO 4, DECEMBER 1997
`
`ROBUST MODEM AND CODING TECH
`
`UES FOR FM HYBRID IBOC DAB
`
`I
`
`an Kroeger, Denise Cammarata
`Westinghouse Wire
`930 Intern
`Linthicum, MD 21090
`
`sidebands contain useful information that is not cbmpletely
`corrupted by an interferer, then CPC codes provlde additional
`
`channel
`FM Hybrid IBOC DAB Parameters
`
`FM lBOC DAB
`
`brief description of the
`etails of the FM BOC system
`ce the presentation
`
`1s
`
`a slope of about -0.35 dB/kHz from the center
`adjacent FM signals, If present, would be cqntered at a
`spacing of 200 kHz.
`Although FM channel
`cing in some countnes is
`100 kHz, these first adjacents
`geographically separated
`such that FM reception is not impaired within the coverage
`area. Therefore this should pose no problem to the FM IBOC
`system. The DAB to DAB interference at 300 kHz spacing
`can impair performance on one sideband, but the CPC code is
`designed to tolerate this condition.
`OFDM
`In the baseline FM IBOC design
`subcarriers are placed on each side of the host FM signal
`occupymg the spectrum from about 130 lcHz throu
`away from the host FM center frequency as sho
`1.
`
`I
`
`1
`
`Figure 1. Power spectral densities of FM and DAB signals
`below FM spectral mask.
`
`I. INTRODUCTION
`
`The focus of this pa
`
`Engrneering Conference
`Forward error
`
`published puncturing techniques [3,4,5] The CPC code
`technique allows the individual transmissions to be combined
`to form a more powerful code than the sum of the individual
`transmissions
`Il3OC DAB is an ideal candidate for the application
`of CPC codes since the digital DAB transmission is
`accomplished over two sidebands (upper sideband and lower
`sideband) which are potentially
`impaired by nearly
`If one
`independent interferers with independent fading
`sideband is completely corrupted by a strong first adjacent
`FM signal in the vicinity of the receiver, the opposite
`sideband must be independently decodable at the receiver.
`coded with an
`Therefore each sideband must be
`independently decodable FEC code. However, when both
`
`Publisher Item Identifier S 0018-5316(57)09255-X
`
`Petitioner Sirius XM Radio Inc. - Ex. 1009, p. 1
`
`

`

`The total DAB power in each sideband is set to about
`-25 dB relative to its host FM power. The individual OFDM
`subcarriers are QPSK modulated at 689.0625 Hz (44100/64)
`and are orthogonally spaced at about 726.7456055 Hz
`(44100” 135/8192) after pulse shaping is applied (root raised
`cosine time pulse with 7/128 excess time functions as guard
`time). The potential subcarrier locations are indexed from
`zero at the FM center frequency to plus or minus 275 at the
`edges of the 400 kHz bandwidth. The outside assigned
`subcarriers are at plus or minus 274 with a center frequency
`of plus or minus 199128 Hz. The inside information bearing
`subcarriers of the baseline system are located at plus or minus
`179 with center frequencies of plus or minus 130087 Hz. The
`pilot subcarriers are located at plus or minus 178 with center
`frequencies of plus or minus 129361 Hz. The subcarriers are
`differentially coded across frequency using the inside pilot
`subcarriers as the reference for the first differentially detected
`symbol. These reference (pilot) subcarriers are modulated
`with an alternating sequence to permit assistance in frequency
`and symbol timing acquisition and tracking. Recent
`evolution of the design shows that adequate acquisition and
`tracking performance is achievable without the pilots.
`
`single blt stream
`
`Required for FMlDAB Wending
`(FM data path not shown)
`
`I
`
`I
`
`deallocate bt
`
`I
`
`I
`sync w r a
`detect
`
`edltting wit k implemented
`as puncturing a code
`
`Erasing ancVor interference suppression
`
`RF receive, digtize, carrier sync,
`symbol sync. mtched filter, detect
`
`Figure 2. Functional diagram showing the mapping and
`processing of bits through the receiver, deinterleaver, and
`FEC decoder.
`
`In the presence of adjacent channel interference, the
`outer OFDM subcarriers are most vulnerable to corruption,
`and the interference on the upper and lower sidebands is
`independent. Since the PSD of an FM broadcast signal is
`nearly triangular, then the interference increases as the
`OFDM subcarriers approach the frequency of a first adjacent
`signal. The coding and interleaving are specially , tailored to
`
`413
`deal with this nonuniform interference such that the
`communication of information is robust.
`The IBOC DAB system will transmit all the digital
`audio information on each DAB sideband (upper or lower) of
`the FM carrier. Although additional subcarriers beyond the
`baseline system can be activated to enable the transmission of
`all the code bits of the rate 1/3 FEC code, the baseline system
`employs a code rate of 2/5. Each sideband can be detected
`and decoded independently with an FEC coding gain
`achieved by a rate 4/5 (optionally rate 213) convolutional
`code. This redundancy permits operation on one sideband
`while the other is corrupted. However, usually both sides are
`combined to provide additional signal power and coding gain
`commensurate with a rate 2/5 (optionally rate 1/31 code.
`Furthermore special
`techniques can be employed
`to
`demodulate and separate strong first adjacent interferers such
`that a “recovered” DAB sideband can supplement the
`opposite sideband to improve coding gain and signal power
`over any one sideband.
`A simplified functional block diagram of the flow of
`the demodulated bits in an FM IBOC receiver is shown in
`Figure 2.
`
`II. CSI AND ADAPTIVE WEIGHTS
`
`Soft-decision Viterbi decoding with (near) optimum
`soft-decision weighting for maximum ratio combining (MRC)
`for differentially detected QPSK subcarrier symbols is
`employed to minimize losses over the channel. Since the
`interference and signal levels vary over the subcarriers
`(frequency) and time due to selective fading, timely CSI is
`needed to adaptively adjust the weighting for the soft-
`symbols. The CSI estimation technique should be designed to
`accommodate a fading bandwidth of up to 13 Hz for typical
`vehicle speeds in the FM band around 100 MHz.
`An expression for the weighting factor can be
`derived assuming gaussian noise into a differential QPSK
`detector resulting in non-gaussian statistics at the output.
`The fading factor can be computed as a function of the
`statistics of the output of the differential detectur where we
`define the soft decision of the form
`S = (a+n,).(a.e.”@ +n,)
`(1)
`where 4 denotes the phase information imposed between a
`pair of adjacent symbols in the differential encoding, and n
`are the independent noise samples. The fading factor a of
`the adjacent symbols is assumed to be approximately equal.
`The signal to noise ratio after differential detection is easily
`computed to be
`
`a4
`SNR =
`2.a’ - 0 2 +04
`The ideal weighting factor for the post-differentially detected
`symbols is therefore
`
`a2
`W =
`2.a2 .cr2 +cr4
`
`(3)
`
`Petitioner Sirius XM Radio Inc. - Ex. 1009, p. 2
`
`

`

`414
`
`The first dflerential approach described here uses
`cal estimates of the second and fourth moments of the
`
`where U,
`n
`
`is the
`
`ng coefficient of the
`
`second and fou
`
`is the square root of equation (3).
`
`and
`
`se can be estimated as
`
`were performed
`
`detection are explored here.
`For moderate to high SNR,
`
`Although long-term estimates without fading yielded good
`ed between long filter
`versus short filter time
`
`hme constants for a
`
`to
`ts were used
`where simple stat
`on confirmed that th~s weight
`estimate 0 2 . Howev
`estimate performed poorly during hmes when the SNR was
`very low due to fading interference
`For example, the
`
`more than the high SNR ap
`
`tion to the weight
`
`errors over
`
`short filter time
`
`nounced when th
`
`nother estimator is
`
`estimate CSI statistics over a
`the estimate should not be s
`
`sufficient accuracy
`significantly less than
`
`technique approximates
`
`and satisfies the
`
`estimated with
`
`differential detechon
`ion of hme (k index) an
`
`3
`
`Petitioner Sirius XM Radio Inc. - Ex. 1009, p. 3
`
`

`

`w k , n =
`
`1
`
`filtd,,, . [ 1 + [ Jiztdk’n
`
`Filter the sequences vk,, and dk,, using second-
`3.
`order digital IIR filters, then compensate for any differences
`in effective group delay to yield sequences filtVk,, and
`fiZtdk,n. The time constant for the j?lhlk,n filter should be
`somewhat smaller than
`the reciprocal of
`the fading
`bandwidth, while the time constant for the filtd,,, filter can
`be somewhat larger. These sequences are representative
`(approximately proportional) of the local mean and standard
`deviation of the sequence v ~ , ~ .
`The sequence of weights for the soft decisions for
`4.
`each subcarrier to be applied prior to differential detection is
`defined as
`
`] ‘1
`
`fiztvk,n - fiztdk3n
`I
`I
`To prevent numerical overflow, check to ensure that
` jilt^^,^ > 1.5-Jiltdk,n in equation (11); otherwise, set the
`weight to zero. Simulation results verified that this weight
`yields good performance under a variety of channel
`impairments with fading and interference.
`
`Smoothing filters for statistical estimates
`
`The values of filtdk and JiltVk,,
`are estimated
`Filtering is
`using filtering techniques described next.
`performed first for each subcarrier at the k& symbol instant in
`time. Then the rows of Jiltd,, and j l t V k , n are simply
`updated across the N subcarriers Equation (12) filters the
`sequences vk,, with a time delay of approximately 16
`symbols, and equation (13) filters the sequences dk,n with a
`time delay of approximately 64 symbols. Both filters have a
`zero frequency gain of nearly unity.
`
`960. subv
`k - 1,n
`
`- 451. ~ u b v
`k - 2 , n
`
`+ 3 . v
`k,n
`
`5 12
`
`(12)
`
`subv
`k,n
`
`=
`
`Additional filtering is performed across the N
`subcarriers. Smoothing the estimates across the N subcarriers
`requires 3 passes of a simple IIR filter. The first pass sets the
`appropriate initial condition of the filter, but does not update
`the estimates. The direction of the second pass is reversed
`from the first, while the third pass is reversed again. This
`results in an approximately symmetric (linear phase) filter
`characteristic which is desirable for providing the estimates
`on the center carrier. Although it is impossible to provide
`
`415
`this symmetric filtering for the subcarriers at each end of the
`band, the impulse response “tails” are folded back into the
`active subcarriers.
`The first pass across the subcarriers sets the initial
`values of filtv,-, and filtd,-, without replacing the time-
`filtered values for each subcarrier. The time index k is
`ignored here since it is understood that the filtering over the
`subcarriers is performed over each k& OFDM symbol.
`
`The second pass smooths the values across the
`+j?.subv,; I
`filtered estimates for each subcanier, S U ~ V and subd.
`I
`Ifiltv, e(l-P)-~Zt~,+,
`jltd, e (1 - p) * filtd,,, + p* Subd, ;
`n= N - 2 , N - 3 , ... 0
`
`(1 5 )
`
`The third pass smooths the frequency values again to
`achieve a nearly symmetrical impulse response (except for the
`subcarriers near the endpoints).
`
`The resulting filtered values for f i b and pltd are used in
`equation (11) at each OFDM symbol time to yield the
`appropriate weight for each soft symbol prior to differential
`detection, but after matched filtering, in the receiver.
`
`IIL CPC CODE FOR OFDM BROADCAST SYSTEM
`
`A simple method of constructing the CPC code for
`this application is to start with an industry standard rate 1/3
`convolutional code. This code can be generated as shown in
`Figure 3.
`The rate 1/3 convolutional encoder of Figure 3 can
`be viewed as producing 3 encoded bit streams (Gl, G2 and
`G3), each at the same rate as the input. The combination of
`these 3 bit streams produces the R=1/3 coded output
`sequence. To create a complementary code pair, for example,
`a subset of the output code bits is assigned to the lower DAB
`sideband and a different (complementary) subset is assigned
`to the upper sideband. Each subset must contain at least the
`same rate of bits as the information input rate, plus some
`additional bits to provide some coding gain.
`The coded bit mask of a Puncture Pattern matrix is
`shown in Figure 4. The Puncture Pattern matrix of represents
`the encoder output symbols over each set of 4 information
`bits. Therefore the output symbols are identified and indexed
`
`Petitioner Sirius XM Radio Inc. - Ex. 1009, p. 4
`
`

`

`416
`
`Fi
`
`atte
`Optionally the punctured bits of the
`be transmitted to yield a pair of rate 2/3 CPC
`as shown in Figure 7. Of course,
`
`codes.
`
`Figure 4. General Puncture Pattern Matrix.
`
`its to achieve maxim
`
`5
`
`bits contribute least to
`e combined code The '
`
`Petitioner Sirius XM Radio Inc. - Ex. 1009, p. 5
`
`

`

`interleaving in frequency such that the erasure of a subcarrier
`would result in periodic erasures at the input to the Viterbi
`decoder [6]. The period of erasures was made sufficiently
`large (i.e., 8) such that it did not reduce the periodic effective
`code length (PECL), which is also bounded by the constraint
`span of the code. Their technique maps adjacent subchannel
`gains to codewords with indexes separated by the period.
`However, Wesel and Cioffi's paper' did not exploit
`interleaving over time to mitigate the effects of flat (or
`wideband) fades over multiple symbol times, nor did they
`exploit the a priori knowledge of nonuniform subchannel
`interference statistics. The latter has resulted in careful
`placement of the code bits over the subcarriers, and the
`selection of the CPC codes for the FM hybrid IBOC DAB
`application.
`Simulation confirmed expectations that, with CSI
`and weighting in both cases, random interleaving performed
`poorly compared to interleaving that results in periodic
`erasures for affected subcarriers.
`Furthermore, careful
`placement of the coded bits using a priori knowledge of
`interference statistics improves performance. The use of CPC
`code techniques as well as interleaving over time further
`improves performance.
`A 255 row by 456 column interleaver array is
`established to hold the bits produced by the convolutional
`encoder. Each row of the interleaver array holds the code bits
`to be modulated in a parallel OFDM symbol. The 256* row
`is reserved for the modem frame sync word. Each pair of
`columns is assigned to the inphase and quadrature QPSK
`modulation of one of the 228 subcarriers. Additional
`subcarriers outside
`the
`interleaver may be used
`for
`transmission of the pilot or other data applications. Code bits
`are written into the interleaver array in a particular pattern.
`The array is read, row by row, providing the data source for
`the parallel OFDM symbols. A pictorial diagram of the
`interleaver array is presented in Figure 9.
`
`38 COLUMNS
`
`INTERLEAVER SIZE
`15 BLOCKS IN 12 PARTITIONS
`= 255 ROWS BY 456 COLUMNS
`
`0
`
`1
`
`2
`
`3
`
`4
`
`PARTITION
`5
`6
`7
`8
`
`9
`
`1
`
`0
`
`1
`
`1
`
`
`
`,
`I
`
`,
`I
`
`,
`I
`
`,
`I
`
`,
`I
`
`,
`I
`
`,
`I
`
`,
`I
`
`,
`I
`
`,
`I
`
`t
`
`~
`
`I
`
`0
`1
`2
`3
`4
`5
`6
`BLOCK 7
`8
`9
`10
`11
`12
`13
`141
`
`Figure 9. Interleaver array.
`
`417
`
`I
`
`partition-assignment =
`
`11 6 0 51
`
`partitions=(O 1 2 3 4 5 FM 6 7 8 9 10 11)
`
`Figure 10. Interleaver partition index assignments.
`
`first
`implemented by
`interleaver can be
`The
`assigning the code bits (modulo 12 index) of the puncture
`pattern to the 12 subcarrier column partitions. T h s is
`illustrated in Figure 10 using the partition index to identify
`the interleaver partitions corresponding to puncture pattern
`bits. The ordering ranges from 0 through 11 over the lower
`frequency subcarriers to the higher frequency subcarriers to
`represent the 12 subcarrier partitions.
`Each partition is comprised of 38 columns and
`carries code bits intended for 19 subcarriers, where the real
`and imaginary components of a particular subcarrier are
`identified as separate adjacent columns.
`The entire
`interleaver consisting of 12 partitions has 456 columns. The
`outermost subcarriers are identified as columns 0,1 and
`454,455. Columns 190 through 265 carry the optional
`punctured bits closest to the FM host spectrum.
`
`I
`
`105 120
`90
`75
`60
`45
`30
`15
`0
`570 585 600 615 630 645 660 675 690
`1140 1155 1170 1185 1200 1215 1230 1245 1260
`1710 1725 1740 1755 1770 1785 1800 1815 1830
`2280 2295 2310 2325 2340 2355 2370 2385 2400
`2850 2865 2880 2895 2910 2925 2940 2955 2970
`3420 3435 3450 3465 3480 3495 3510 3525 3540
`3990 4005 4020 4035 4050 4065 4080 4095 41101
`4560 4575 4590 4605 4620 4635 4650 4665 4680
`5130 5145 5160 5175 5190 5205 5220 5235 5250
`5700 5715 5730 5745 5760 5775 5790 5805 5820
`6270 6285 6300 6315 6330 6345 6360 6375 6390
`6840 6855 6870 6885 6900 6915 6930 6945 6960
`7410 7425 7440 7455 7470 7485 7500 7515 7530
`7980 7995 8010 8025 8040 8055 8070 8085 8100
`8550 8565 8580 8595 8610 8625 8640 8655 8670
`9120 9135 9150 9165 9180 9195 9210 9225 9240
`1
`16
`31
`46
`61
`76
`91
`106 121
`Figure 11. A portion of the interleaver array (rows 0
`through 17, and columns 0 through 8) showing the
`spacings of the kth partition index.
`
`6
`
`Petitioner Sirius XM Radio Inc. - Ex. 1009, p. 6
`
`

`

`418
`
`Each partition is further divided into 15 blocks of 17
`rows each. These blocks facilitate the interleaving over time
`onding to adjacent coded
`by separating code bits, CO
`information bits, by the number of rows in a block
`The interleaver array row and column indexes, TOW
`
`I
`
`I
`
`m
`
`e
`
`constants
`
`of the kth puncture
`
`interleaving and espec
`
`V. SlMULATION RESULTS
`
`Some assumptions and simulation conditions are
`
`* - -
`
`'
`
`__ .. . .
`
`. . .
`.
`
`
`.
`
`-. . . .
`
`I - I O - ~
`
`$
`m
`
`I
`
`I
`. !
`
`I
`
`
`
`I
`
`\
`
`Signal-to-noise ratio is represented as E,& which is the ratio
`of the energy of one channel code bit to the noise in a 1 Hz
`
`I
`
`'
`
`
`
`I
`
`10
`
`15
`
`'
`
`8
`
`
`
`I
`
`5
`
`1 - 1 0 P 1
`0
`- AWGN with no interferers
`E d "
`- - -
`AWGN, one 1st adj FM at -6dB (w/ FAC)
`AWGN, two 1 st ad1 FM at -6dB (w/ FAC)
`
`which effectively reduces th
`first adjacent FM signal
`
`channel scenarios is shown in Figures 12 and 13 Edge of
`
`7
`
`igure 12. BER performance in AWGN without selective
`iding under various interference conditions.
`
`The third curve of F'
`case scenario involving two -6
`the same receiver location. This curve is much flatter than
`
`Petitioner Sirius XM Radio Inc. - Ex. 1009, p. 7
`
`

`

`the other two since both DAB sidebands are impacted. Virtual
`CD quality is still achievable at 13 dB which is the SNR at
`the edge of coverage given a equivalent thermal noise
`temperature of 60,000 degrees K.
`The second series of performance curves is shown in
`Figure 13 represent performance in the mobile environment.
`Two different multipath models were used. The parameters
`of the 9 ray model are from EIA’s “Urban Fast Rayleigh
`scenario 171. Delays range from 0 to 3 microseconds and
`attenuations are from 2 to 10 dB; the fading bandwidth is 5
`Hz. The second model places more nulls across the DAB
`bandwidth and has faster fades (10 Hz). Three rays are used
`with delays of 0, 5, and 20 microseconds and no relative
`attenuation. Two scenarios were run with each selective
`fading model - one with a -6dB FM first adjacent and one
`without adjacent interference. The slower fading offered the
`greater challenge to DAB performance, but results were still
`excellent. Virtual-CD quality was again achieved in all cases
`assuming an equivalent noise temperature of 40,000 degrees
`K, generally leaving ample margin in the coverage area.
`
`419
`Techniques for estimating the optimum soft-decision
`weight for QPSK symbols prior to Viterbi decoding were
`described. These techniques apply to differential detection of
`multi-carrier (OFDM) QPSK, with and without multipath
`fading of the signal of interest or the interferer. The fading
`cases can necessitate a compromise between accurate CSI
`estimation and agility of the CSI to track the fading signal or
`noise components.
`Robust approximate estimation techniques were
`found for estimation of the weight in the presence of
`independently faded signal and noise. This robust technique
`is a result of the compromise between accurate CSI estimation
`and agility in fading.
`Simulation results presented here confirm the robust
`performance of the FM hybrid B O C system, even in the
`presence of a strong first-adjacent interferer. Virtual-CD
`audio quality should be generally achievable within and
`beyond the normal FM coverage area.
`
`ACKNOWLEDGEMENTS
`
`I
`
`U U I
`
`___
`I
`
`I
`
`-
`
`I
`
`I
`
`__ __ _. - -. ____
`n
`
`I
`7
`I
`
`I
`
`I
`
`The authors gratefully acknowledge the expert
`assistance of Dr Carl-Enk Sundberg of Lucent Technologies
`and Brian Clien of MIT (on summer assignment to Lucent)
`for identifying the optimal puncture patterns and their
`corresponding free distances and weights for our CPC code
`design.
`
`REFERENCES:
`
`[l] B. Kroeger, P. Peyla, “Robust IBOC DAB AM and FM
`Technology for Digital Audio Broadcasting,” 5 lSt Annual
`Broadcast Engineering Conference (NAB), Las Vegas,
`Nevada, April 1997.
`[2] S. Kallel, “Complementary Punctured Convolution
`(CPC) Codes and Their Applications,” IEEE Trans.
`Comm., Vol. 43, No. 6, pp. 2005-2009, June, 1995.
`[3] J. Cain, G. Clark, Jr., & J. Geist, “Punctured
`Convolutional Codes of Rate (n-l)/n and Simplified
`IEEE Trans.
`Maximum Likelihood Decoding,”
`Information Theory, Vol. IT-25, pp. 97-100, Jan. 1979.
`[4] Y. Yasuda, K. Kashiki, Y. Hirata, “High-Rate Punctured
`for Soft Decision Viterbi
`Convolutional Codes
`Decoding,” IEEE Trans. Comm., Vol. 32, #3, Mar. 1984.
`151 J.
`Hagenauer,
`“Rate-Compatible
`Punctured
`Convolutional Codes
`(RCPC Codes) and Their
`Applications,” IEEE Trans. Comm., Vol. 36, No. 4, pp.
`389-400, April, 1988.
`[6] R. Wesel, J, Cioffl, “Fundamentals of Coding for
`Broadcast OFDM,” Proceedings of the 29& Asilomar
`Conf. on Signals and Systems, 1995.
`[7] T, Keller et. al., Digital Audio Radio Laboratov Tests
`Transmission Quality Failure Characterization and
`Analog Compatibility Vol. I , Section E, pg. 2, EIA,
`Consumer Electronics Group, August 11, 1995.
`
`I
`4
`
`I
`6
`
`I
`8
`
`I
`12
`
`I
`14
`
`I
`16
`
`I
`I O
`EsNo
`
`- Three ray selective fading w/ no interferers
`- -
`-
`
`Nine ray selective fading w/ no interferers
`Three ray sel. fading wi -6dB 1st adj. FM (wlFAC)
`Nine ray sel. fading w/ -6dB 1st adj FM (wlFAC)
`
`Figure 13. BER performance with selective fading under
`various interference conditions.
`
`VL CONCLUSIONS
`
`The FEC code and corresponding interleaver design
`for an F M IBOC system using OFDM was presented here.
`The existence of a good-performing pair of rate 4/5 CPC
`codes has been proven- Analysis and simulation has shown
`that careful mapping of the modulated code bits chosen from
`a puncture pattern to the. OFDM subcarriers can mitigate the
`effects of potential first-adjacent interferers.
`
`8
`
`Petitioner Sirius XM Radio Inc. - Ex. 1009, p. 8
`
`

`

`420
`
`articipates in the analysis
`simulation of DAB
`rithms and leads the
`
`m, Denise spent the last
`
`Technology Council. Denise holds a Master of Science (MS)
`in Electrical Engineering from T
`sity and a Bachelor of Science degr
`from Loyola College in Baltimore, Mary1
`
`accomplishme
`His
`communications earned him
`
`t courses
`
`in
`
`Innovation Awards in previous years, in addition to
`approximately 70 other technical awards,
`
`9
`
`Petitioner Sirius XM Radio Inc. - Ex. 1009, p. 9
`
`