`Samsung v. Affinity
`IPR2014-01181
`Page 00001
`
`
`
`on
`
`Rm : -[Pym _l'0_o.3s.lf|.:'.o .af : 2412747. ppm
`—u:
`
`Then the pea]; of the FM power spectral density-is located
`13.8 dB ('10- l0g(24.12'?47) 1 below the total carrier power
`reference level (0 dB), as shown in Figure 1. The DAB
`power level on each side of the FM spectrum is placed 25 dB
`below the total FM power (this value. is adjustableby the
`broadcaster to accoimuodate special interference situations}.
`The DAB density in a 1-kHz bandwidth can be calculated.
`The power spectral density of the DAB signal can be very
`closely approximated by dividing its total power by its
`effective Nyquist Bandwidth.
`0 .25: I n
`= 31-0.796375
`Then the power spectral density of the DAB signal in dB. as
`shown in Figure
`1,
`is computed to be -43
`dBt'kHz
`
`: 4,9 - 10*
`
`PSDM,
`
`(10- log(4.9- 1 0-5)).
`
`l'(I.0.062S§ 0.5
`
`Figure 2. Plot showing rectangular Nyquist pulse (dotted)
`and the root raised cosine tapered pulse -(solid).
`
`The baseline DAB system assttrncs 95 sub'carricrs
`above and 95 below the host FM spectrum.
`Each DAB
`subearrier
`is QPSK modulated at
`a
`symbol
`rate of
`approximately 689 Hz. The inphase and quadrature pulse
`sliapes are root-raised—eosine tapered (excess t'urie=7t 128) at
`the edges to suppress the spectral sidelobes. Although this
`pulse. shape reduces the throttghput capacity relative to the
`rectangular pulse by 5.5%, performance in nutltipadi
`is
`improved and the resulting spectral sidelobes are reduced,
`lowering interference. This pulse shape results in orthogonal
`subcarrier frequency spacing of approximately 727 Hz. A
`plot of the pulse shape normalized to -.1 -ui:Iit'_c_Il' time is
`presented in Figure 2. Figure 3 shows-plots of one-of the t-wo
`sidebands of the DAB spectra using both the rectangular and
`root-raised-cosine pulse shapes.
`
`D.-*\B(kH:!..0)
`
`Dr\Btkl[z_.0.t}G25}
`
`sidelobe su'pp_I'ession of
`Improved spectral
`Figure 3.
`Nyquist
`root
`raised cosine tapered pulse (solid) and
`rectangular Nyquist pulse (dotted).
`
`Potential subcarrier locations are identified by their
`ofiset from the host FM center frequency. These subcanicrs
`are indexed from zero at the-center FM frequency to plus or
`minus 2?3 near
`the edges of the 400 kHz bandwidth.
`Subcatriers 179 through 273 correspond to 13008? Hz to
`198402 Hz= from the center. These subcarriers carry about
`105 kbps of information which is error protected with a rate
`2X5 FEC. This data rateis sufficient for Lranstttisstort of
`virtual-CD quality music plus a modest datacastitrgcapacity.
`Optionally, additional carriers can be added to increase the
`datacasting capacity. These carriers would be located closer
`to the host analog FM signal.
`.
`I The placement of DAB at i 15 kHz about 114 kHz
`is-avoided in the baseline system in order to reduce the noise
`introduced into inadequately filtered receivers. However the
`broadcaster will have the option to utilize this portion of the
`spectrum to. improve 1'obustness of the digital audio- signal
`ztndfor to provide additional datacasting-capacity.
`
`FEC CDt:l.ir1g
`Forward error correction and interleaving improve
`the reliability of the transmitted information. In the presence
`of adjacent channel interference, the outer OFDM subcaniers
`one rnost xnrlneratuc to corruption, and tilt: 1ntt:rt'r:r'cnce on the
`upper and lower sidebands is independent. The infonnation.
`coding-, and interleaving are specially tailored to deal with
`this‘nonu'1'u'form interference such that the corumurucalion-of
`ittforrttation
`is
`robust.
`Specifically,
`this nonunjforrn
`interference is thefocus here where special coding and error
`haiidl'ttt'g results in more robust performance.
`.
`The [BOC DAB system will transmit all the digital
`audio ‘information on each DAB sideband (upper or lower] of
`' the FM carrier. Recall that the baseline system constrains the
`
`(cid:51)(cid:68)(cid:74)(cid:72)(cid:3)(cid:19)(cid:19)(cid:19)(cid:19)(cid:21)
`
`
`
`DAB signal to within approximately 130 kHz to 199 kHz
`above and below the FM center frequency, as shown in Figure
`1. Each sideband can be detected and decoded independently
`with an FEC coding gain achieved by a rate-4z'5 convolutional
`code on each sideband. 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. Furthermore. special techniques are
`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.
`The goal here is to transmit the DAB signal on both
`the upper and lower sidebands such that the sidebands can be
`independently detected and decoded, each with some FEC
`coding gain. Additional coding gain, along with some power
`gain of course,
`is desired when both sidebands can be
`combined. The reason for these requirements is that
`the
`interference on each sideband is independent of the other;
`however, the level of interference across the subcarriers on
`any one sideband is related to the power spectral density of
`the
`adjacent
`interferer.
`Therefore
`the grouping
`of
`independently
`detectable
`and
`decodable
`sidebands
`is
`appropriate.
`In order to effectively achieve coding gain when the
`pair of sidebands is combined,
`the code on each sideband
`should consist of a subset of a larger (lower rate) code. Each
`subset can be designed through “complementary” puncturing
`of the lower rate code. Details of the FEC design are
`presented in [2].
`
`HI. INTERFERENCE ANALYSIS
`
`The interference to and from the first adjacent
`channels placed i 200 kHz from the host signal can be
`derived from the relationship of the adjacent signals shown in
`the plot of Figure 4. FM stations are geographically placed
`such that
`the nominal
`received power of an undesired
`adjacent channel is at least 6 dB below the desired station's
`power at the edge of its coverage area. Then the DIU {desired
`to undesired power ratio in dB) is at least 6 dB. Knowledge
`of the ratio of each stau'on’s DAB signal power to its FM host
`permits assessment of first adjacent
`interference to DAB.
`Similarly the interference of the first adjacent DAB to the
`host FM signal can be assessed from the relationship.
`Figure 5 illustrates the need for DAB spectral
`sidelobe suppression and bandlimiting due to the second
`adjacent DAB interference to the host DAB signal. At a
`station's edge of coverage. a second adjacent’s nominal power
`can be up to 20 dB greater than the host’s nominal power.
`The effects of the various interference scenarios
`
`illustrated here are quantified through analysis and supported
`through simulation and testing. Analysis of the DAB to first
`adjacent interference at the edge of coverage showed that the
`total DAB signal should be set about -22 dB relative to its FM
`power.
`
`Fli-'lI[}AB(kH2.0,0}
`
`FivlD.-1B(k.Hz,2fl0,—6)
`
`Interference scenario showing first adjacent at
`Figure 4.
`-6 dB (worst case edge of coverage).
`
`Flvl]1|AB(kI-lZ.'Z|,l))
`
`£.1]I)A]3(kI-12,400.20)
`
`Figure 5. Interference scenario with second adjacent at
`+20 dB.
`
`interference
`Iirst adjacent
`The solution to the
`problem is to place redundant although not identical, DAB
`signals on either side of the carrier. Although the potential
`capacity is halved with this
`redundancy,
`interference
`problems are substantially reduced and substantial coding
`gain is achieved after combining both halves. A survey of
`existing U.S. radio allocations shows that it is very unlikely
`that both upper and lower adjacent channel interferers are
`present at their ntaximutn interference levels (-6 dB} at the
`same geographic location within the host’s coverage area.
`This frequency diversity is especially useful when multipath
`interference or spectral notches affect one sideband or the
`other.
`
`have
`analyses
`and
`simulations
`of
`A variety
`in the
`characterized performance of the host FM signal
`presence of IBOC DAB. Specifically. main audio channel
`performance, SCAS, adjacent channels, and stereo subcai-n’cr
`demodulation were investigated with an IBOC DAB signal
`appended to the host FM.
`
`(cid:51)(cid:68)(cid:74)(cid:72)(cid:3)(cid:19)(cid:19)(cid:19)(cid:19)(cid:22)
`
`
`
`Alain rmdio charmel perforrnance
`Simulations have provided valuable insight into the
`character of W post-detection noise in the presence of an
`]ZBO_C DAB signal. For instance, results indicate that the
`audio noise level
`increases with the deviation of tile FM
`signal.
`In fact. Figure 6 illustrates a significant rise in the
`post-detection noise power spectral density (PSD) as the FM
`deviation varies from minimum to maximum in the presence
`of an IBOC DAB signal placed between 78 kHz and 197 kHz
`from the FM carrier.
`
`frequency-division multiplexing (OFDM) with 475.0-symbol-
`per-second quadrature phasc—shift keying (QPSK) subcarriers
`using rectangular pulse shaping. The filth test employed
`DAB with four times the number of 0FDM.carIier_s -- each
`Occupying one-fourth the bandwidth (11875 Hz) -- and root-
`raised-cosine pulse shaping (to reduce spectral sidelobes that
`interfere with the host FM}.
`In each test, Lhe- spectral
`occupancy of the DAB signal was chango_d:_
`the start
`frequency was varied with respect to the FM center frequency,
`while the stop frequency was fixed at 197 kHz. Table l
`surrtmarizes the results.
`.
`
`.
`
`Post-detection Noise Floor (78~kI-lz DAB‘)
`
`0 I i
`
`Table 1 - Audio Dynamic Range at Transmitter
`'
`cak-to-noise-floor SNR
`
`DAB start frequency
`
`Audio SNR
`dB)‘ 15 kltl-2.
`
`.
`
`Frequency (kl-lz}
`""" Near peak deviation
`' ' Minimum deviation
`
`Figure 6. Audio Deviation Effects.
`
`responsible for
`is
`The nonlinear FM detector
`internlodulating overlapping portions of the host FM and
`DAB spectra. The products are folding back into the ‘post-
`delection audio band and raising its noise floor.
`Similar
`observations and conclusions were independently reached by
`the Electronic Industries Association (ELA) during their
`moo DAB testing. [4].
`'
`Although these results are intriguing, they do not
`predict a degradation in host FM audio quality due to IBOC
`DAB. Because the DAB-induced post-detection noise floor
`increases in proportion to the deviation of the FM signal, the
`effect is self-masking: audio noise will be lowest during quiet
`passages,
`and highest only when the audio is
`loudest.
`Simulations have demonstrated this phenomenon.
`The absolute level of host PM degradation will
`depend on the particular configuration of DAB.
`‘l‘o
`deter-mine the relationship between DAT} location and audio
`signal-to-noise ratio {SNR), a number of perforruaner; tests
`were run when DAB noise would "be most audible —- during
`quiet passages of minimum FM deviation. Simulations were
`performed in which the receiver audio
`range was
`rueasuied witlr only a 10%-deviated, l9-kHz:-pilot-modulated
`FM signal and a DAB signal input to an FM stereo receiver
`located at the transmitter. The total power of the DAB signal
`was 22 dB‘ below the power of the W _carricr.
`In the first
`four
`tests, Lhe DAB was modulated using onliogonal
`
`These results indicate that moving the‘ DAB away
`from the FM carrier, increasing the number of DAB carriers,
`and pulse shaping the transmitted DAB symbols to reduce‘
`spectral sidelobes will significantly improve the perionnance
`of the host FM. Modulation and coding characteristics of the
`DAB signal can be traded for spectral occupancy; to meet
`these goals.
`Note that
`the new DAB baseline ‘employs
`subcarriers
`spaced at
`'.l2'r', Hz which would improve
`performance over the carrier spacings reported here.
`Audio simulations have verified that an SNR of 77 .6
`
`dB during quiet passages should render DAB-induced audio
`noise
`imperceptible
`to
`the
`listener.
`Furthermore,
`implementation constraints limit the SNR oftypical receivers
`to around 60 "dB. The" noise engendered by these freceivers
`will mask any degradation caused by DAB. The -22-.dB. 129-
`kl-lz pulse-shaped DAB configuration is used as the‘ baseline
`for the balance of this "discussion.
`I
`
`S04 p.e-rfortrim
`.
`SCAS (Subsidiary Conrrnunications Authorization)
`are optional channels multiplexed onto the baseband stereo
`spectnnu from 53 kHz to 100 kHz. The SCA signal, which
`can be analog or digital, is transrnitted by some FM stations
`for the use of private subscribers who typically pay for
`program material. Simulations were used to determine the
`impact of SCAs on IBOC DAB host FM perfonuartce, and to
`determine the impact of DAB on the performance of SCAs.
`SCAs V.‘tr‘lti’l 10% deviation at 67 kHz and 92 kHz were
`simulated because they represent
`a
`large percentage of
`operational subcarriers.
`In the current analog FM system, SCAS generally
`cause negligible interference to the host FM signal. However,
`when DAB is present, the addition ‘of SCAS could increase
`
`(cid:51)(cid:68)(cid:74)(cid:72)(cid:3)(cid:19)(cid:19)(cid:19)(cid:19)(cid:23)
`
`
`
`to the DABi'FM
`the host PM audio noise floor due
`interrnodulation effect described above. Figure 7 illustrates
`stereo subcarrier sensitivity to 92-kl-Iz SCAs when subject to a
`pulse-shaped (PS) DAB signal starting at 129 kHz.
`ln this
`case, the 92~kHz SCA reduces the host FM audio SNR from
`77.6 to 69.8 dB; however. this noise level is still too low to
`produce audible effects. Figure 8 shows that SCAs located at
`67 kHz have even less impact on audio performance.
`
`Post-detection Noise (I29-kHz PS DAB)
`
`425
`
`the location of the pre-detection DAB).
`this is nearest
`Moreover. the noise masking effect described above does not
`apply for SCAs, since their audio may be quiet while the
`main audio channel, at peak deviation. is causing an increase
`in the SCA noise floor.
`
`Simulations were perfonned using SCAS with peak-
`deviated audio signals in the presence of a -22-dB, 129-kl-lz
`pulse-shaped DAB signal. Figure 9 indicates that the SNR of
`a 6?-kHz SCA (in a 10-kHz bandwidth) is 25-30 dB at the
`transmitter when the main audio channel is near maximum
`deviation.
`
`Post-detection Noise (129-kHz PS DAB)
`
`so
`
`Frequency (Id-I2)
`" 92-kHz SCA, minimum deviation
`‘ ' No SCA, minimum deviation
`
`Figure '7. Effects of 92-kHz SCA
`
`30
`60
`Frequency (Id-lz)
`
`100
`
`120
`
`— 67-kHz SCA, near peak deviation
`
`Post-detection Noise (129-kl-lz PS DAB)
`
`Figure 9. 67-kHz SCA Performance.
`
`For 92-kHz SCAS,
`illustrated in Figure 10.
`
`the SNR is 20-25 dB,
`
`as
`
`Frequency (kl I2)
`"“ 67-kl-Iz SCA, minimum deviation
`‘ ' No SCA, minimum deviation
`
`Figure 8. Effects of 67-kHz SCA.
`
`Due to their location at the high end of the basehand
`spectnim, some SCAS currently operate at low SNRS because
`the post-detection noise floor increases with the square of the
`frequency. When DAB is added, the deviation of a wideband
`host FM signal
`into its
`IBOC DAB signal produces
`intermodulatjon which increases the post-detection noise
`floor, particularly in the higher baseband frequencies {since
`
`20
`
`40
`
`6t]
`
`30
`
`l|'.l0
`
`Frequency (kHz)
`
`* 92-kHz SCA, near peak deviation
`
`Figure 10. 92-kHz SCA Performance.
`Without DAB,
`typical noise floors are roughly 40
`dB, The increase in noise floor should not pose a problem for
`digital SCAS (e.g., Seiko and Radio Broadcast Data System).
`
`(cid:51)(cid:68)(cid:74)(cid:72)(cid:3)(cid:19)(cid:19)(cid:19)(cid:19)(cid:24)
`
`
`
`426
`
`since they should be robust enough to operate at reasonably
`low SNRs.
`
`/ldfacent clranrrei Qergomiance
`The Federal Conimunications Commission defines
`the “edge of coverage” for class B stations as the 54-dBu
`coveragecontour, where 0 dBu‘is equivalent to 1 rnicrovolt
`per meter field strength. Class B stations are protected to the -
`S4-dBu contour from 48-dBu interference introduced by first
`adjacent stations. ' Thus. at
`the edge of coverage, ' the
`minimum desired—to-tutdesired signal ratio (D.-’U]
`is 6 dB.
`Simulations have quantified the amount of degradation that
`would be introduced into the desired IBOC DAB host FM
`
`signal when located at the edge of coverage and subject to
`interference from a -6 dB IBOC DAB first adjacent.
`it is first
`To properly interpret simulation results.
`necessary to calculate the audio SNR of the simulated receiver
`at the 54-dBu contour, assuming that the noise contribution is
`due solely to ambient additive white gaussian noise (AWGN).
`Assuming a mid-baud carrier frequency of 100 MHz and a
`half—wave dipole antenna, electric field intensity E (Wm) can
`be converted to carrier power C (W) at the input to the FM
`receiver using
`
`where A. = 1.17? m2 is the eflective aperture of the half-wave
`dipole antenna. Using this formula, a 54 dBu field strength
`corresponds to a -91.1 dBW carrier power.
`An ambient noise temperature of 10,000 K is
`representative of the FM frequency band;
`in a 15-kHz
`bandwidth, this temperature produces a noise power of -146.8
`dBW. Hence, a carrier power of -91,! dBW at the receiver
`antenna terminals would yield a 55.7 dBf15 kHz carrier-to-
`noise ratio {CNRJ. The receiver noise characteristic enables
`one to detertrtine audio SNR given an input CNR. Using the
`measured noise characteristic of the siniulated FM stereo
`receiver, this input CNR corresponds to an audio SNR of 64.4
`dB! 15 kHz.
`
`'
`
`Recall that a -22-dB pulse-shaped IBOC DAB signal
`starting at 129 kHz yields an audio SNR of 'l7_'6 dB at the
`transmitter {and at the edge of coverage if ambient noise were
`ignored}. The preceding calculations demonstrate that, at the
`edge of coverage, the contribution to audio SNR is dominated
`by ambient noise;
`the effects of —22—dB. 129-kHz pulse-
`. shaped IBOC DAB are negligible.
`In the simulation, a -22-dB pulse-shaped DAB Signal
`starting at 129 kHz was added to both a quiet ‘Fh/I host at the
`transmitter (10%-deviated I9-kHz pilot, with no audio or
`SCAS) and a -6 (13, fully modulated first adjacent. A 150-
`kHz prc-detection filter (300-k.Hz total 3-dB'bandwidllt) was
`used in the simulated FM stereo receiver. The signal at the
`input to the FM demodulator is shown in Figure I 1,
`Results indicate that introduction of the adjacent
`LBOC DAB channel degrades the audio SNR to 50.0 dB.
`Altltough the simulation was perforated with the desired
`
`the
`that
`is clear
`it
`transmitter,
`its
`located at
`signal
`introduction of noise associated with- translation to the edge of
`coverage would be negligible. Therefore, when a -6-dB first-
`adjacent FIWDAB signal impinges on a quiet host-FMIDAB
`signal at the edge of coverage, the SNR degrades from around
`64 dB to 50 dB. Figure 12 illustrates this elIect_
`'
`
`Pre—detection FM {I50-kH.z filter)
`
`so
`"1692so—2oo—'1so-too -50 n
`Frequency (kHz)
`
`too 150 zoo 250
`
`* FM detector input
`Figure 11. Pre-detection Effect of First Adjaeentlwith
`I29-kflz PS DAB at edge of coverage.
`
`Post-detection Noise (150-l-{Hz filter)
`
`60
`
`80
`
`100
`
`120
`
`Frequency (kl iz)
`
`‘—' 129'-kHz PS DAB with first adjacent
`' '
`129-kHz. PS DAB at edge of coverage
`
`Figure 12. Post-detection Effect ofFirst Adjacent with
`129-kHz,PS DAB at Edge of Coverage.
`
`While the SNR is diminished, it should be_ noted that
`degradation
`is
`highly
`geographically
`localized;
`the
`performance will
`improve rapidly as the receiver moves
`farther from the interfering station or closer to the desired
`Station.
`In addition, due
`to
`receiver
`impleincntation
`constraints, actual receivers will experience less than 64-dB
`SNRS at the edge of coverage. Most automotive receivers are
`blended to mono at the edge of coverage anyway, mitigating
`the effects of first adjacent DAB interference. This would
`
`(cid:51)(cid:68)(cid:74)(cid:72)(cid:3)(cid:19)(cid:19)(cid:19)(cid:19)(cid:25)
`
`
`
`improve audio SNR by removing the effects of noise around
`the stereo subcarrier.
`
`Stereo and carrier demodulation
`
`Dining EIA testing of the USADR FM-I IBOC DAB
`system, certain inexpensive FM stereo receivers suffered an
`increase in audio noise when receiving an IBOC DAB FM
`stereo signal [4}. When the DAB signal was removed from
`the FM signal, the audio noise disappeared.
`Investigations
`revealed that the problem was caused by inadequate filtering
`of the post-detection baseband stereo multiplex signal. The
`new baseline DAB wavefonn has been designed to mitigate
`this effect.
`To recover the stereo information, the 30-kHz-wide,
`double-sideband amplitude-modulated (DSB) left-minus-right
`(L-R] signal centered at 38 kHz is demodulated using a 38-
`kHz local oscillator (L0), and subsequently filtered with a 15-
`kllz lowpass filter.
`In most receivers,
`the 38-kHz L0 is
`simply a square wave, with a 38-kHz fundamental and odd
`harmonics at 114 kHz, 190 kHz, etc. As a result,
`in the
`absence of adequate filtering, not only is the desired L-R
`signal recovered, but so is any energy in the multiplex signal
`that lies within :15 kHz of 114 kHz and 190 kHz.
`In the presence of AWGN only (no DAB), this effect
`is not pronounced. A well-known property of large-signal
`FM detection in AWGN indicates that the power spectral
`density of the post-detection noise is directly proportional to
`the square of the frequency. Hence, the noise power spectral
`density at 114 kHz is 9 times that at 38 kHz (9.5 dB), and the
`noise at 190 kHz has 25 times the power (14.0 dB}. High
`noise levels are mitigated because the amplitude of square
`wave harmonics decreases with their order:
`if the 38-1-LI-Iz
`fundamental has unit amplitude. the I14-kl-Iz third harrnonic
`has amplitude M3 (-9.5 dB), and the 190-kl-I2 fifth harmonic
`has amplitude US (-14.0 dB).
`Therefore.
`the noise
`contribution from each harmonic is equal to the noise under
`the desired signal;
`this causes a 4.8-dB degradation due to
`AWGN alone (without DAB) in receivers which do not filter
`the noise around their L0 harmonics.
`
`This decrease in SNR is avoided in well-designed
`receivers.
`Some receivers use “Walsh” decoders;
`others
`simply filter the bascband multiplex signal prior to DSB
`demodulation, which effectively
`eliminates
`components
`outside the desired L-R band. Most receivers —- even those
`
`without such post-detection protection — should ameliorate
`the effects of the I90-kl-Iz fifih harmonic by pre-detection
`filtering, since a good design would significantly filter the
`first adjacent FM signal centered 200-kHz from the desired
`channel.
`
`in the presence of AWGN alone. certain
`Thus,
`inexpensive receivers which employ little or no post-detection
`protection experience up to a 3-dB stereo SNR degradation
`(from their DSB LE) third harmonic} when compared to their
`more carefully designed counterparts.
`Of course,
`no
`significant degradation exists when receiving a monaural
`signal.
`
`42'!
`
`As discovered during FM-l EIA testing, this 3-dB
`stereo SNR degradation increases when IBOC DAB is added
`to the analog FM signal [4].
`In order to scope the magnitude
`of the problem, simulations were performed using a well-
`designed FM stereo receiver with ample protection from 38-
`kllz harmonics.
`the first simulated
`Three simulations were run:
`performance in a well-designed receiver by adding AWGN
`only to a quiet analog FM signal at a level which produced a
`64-dB SNR in the lefl audio receiver channel. The second
`added DAB only (from 78 kHz to [97 kHz) to the quiet FM
`signal, at a level which likewise produced a 64-dB SNR in the
`left audio channel. As shown in Figure 13, the pest-detection
`noise power in the 0-53-kHz audio band is identical for the
`two simulations (hence the equal 64-dB SNRs}.
`
`Post-detection Noise
`
`Frequcncy(kHz)
`
`1 78-kHz DAB (sim 2)
`- -
`l29—l<Hz PS DAB (sim 3)
`" Typical receiver: 64-dB AWGN (sin: 1)
`
`Figure 13. Effect of DAB on 114-kHz Noise Floor.
`
`In
`Note, however, that the noise floors diverge above 60 kHz.
`fact, the DAB-induced noise floor is approximately 25 dB
`higher in the 30-kl-iz band around 114-kHz. Ifthc simulated
`receiver did not sufficiently filter the post-detection noise
`floor, the stereo noise increase would have degraded the audio
`SNR well below 64 dB in the second simulation.
`
`It has been suggested that simply suppressing DAB
`energy in the 114115 kHz band would eliminate the post-
`detection noise in this region. Due to the non-linear nature of
`the FM demodulator, this is not entirely the case.
`Instead,
`simulations have shown that such a notching of DAB carriers
`creates a 12-dB improvement across the 30-kHz hand around
`:14 kHz. Thus, simply limiting DAB bandwidth might still
`cause stereo SNR degradation in radios with inadequate post-
`detection filtering.
`Significant improvements, however. were observed
`in the third simulation. in which the DAB signal was moved
`beyond 129 kHz and pulse shaping was applied.
`Pulse
`shaping causes a significant decrease in the noise floor across
`
`(cid:51)(cid:68)(cid:74)(cid:72)(cid:3)(cid:19)(cid:19)(cid:19)(cid:19)(cid:26)
`
`
`
`42%
`
`I much of the post-detection band, as illustrated in Figure 13
`(l29~kHz PS DAB). Figure 14 provides a magnified view of
`the 30-kHz region around I 14 kHz.
`'
`.
`
`Post-detection Noise
`
`- 78-kHz DAB (sun 2)
`' '
`129-kHz PSDAB (sim 3)
`" Typical receiver: 64-dB AWGN {sim l)
`
`Figure 14'. Effect of DAB Placement and Pulse Shaping on
`114-kHz Noise Floor.
`
`Note that the noise floor steps up at 110 kHz, due to
`mixing of the Bessel-weighted 19-kHz pilot harmonics with
`the DAB signal during FM demodulation. As a result, above
`110 kHz, an improvement of around 10 dB is gained over that
`afiorded by 78-kl-I7. (non-pulse-shaped) DAB. Below 110
`kHz, however, a 30-dB improvement
`is observed.
`Thus,
`using DAB placement and pulse shaping, the overall stereo
`noise increase due to the addition of DAB in radius with
`
`inadequate filtering can efifectively be limited to acceptable
`levels.
`
`The preceding analysis presents a worst-case bound
`on the 1l4~kl-Iz degradation due to DAB;
`in reality, even the
`most poorly designed receivers should provide some. degree of
`pre-detection filtering to mitigate the noise level around li4
`kHz. Furthermore, in a typical environment, with practical
`receiver implementations, -it is probable that the degradation
`Will be imperceptible to the listener. Note that none of the car
`radios employed in the EIA tests exhibited the problem [4].
`
`interference Stttilfllafl
`on
`impact
`the
`analyzed
`has
`Westinghouse
`performance of the host FM signal in the presence of various
`IIBOC DAB configurations. Simulations and-. analysis indicate
`that FM performance is least affected when a pulse-shaped
`DAB signal is placed between 129 kHz and 197 kHz from the
`FM carrier. Modulation and coding tradeofifs can be
`exercised to provide the spectral elficiency required to fit the
`DAB.signal within this bandwidth,
`This DAB configuration yields an audio SNR of
`nearly 78 dB during periods of minimum deviation, with
`
`noise during louder passages rendered inaudible to the
`listener via a masking effect. Even when quiet, noise due to
`DAB will probably be masked by the noise produced in most
`typical receivers. SCA interference with the host
`should
`likewise be inaudible, while the SC.As themselves should
`perform with SNRS of around 20-30 dBf10 kHz (ample
`margin for a digital SCA}_
`.
`-
`When a high-level first adjacent interferer w§ith'DA}3
`is present, the audio SNR of the host (with DAB) degrades to
`50 dB at the edge of coverage. However, the degradation is
`highly localized. Most automotive receivers are blended to
`monoat the edge of coverage anyway, mitigating the effects
`of first adjacent DAB interference.
`Finally, a slight degradation may be observed during
`stereo snbcarrier demodulation in "existing inexpensive FM
`stereo receivers.
`This degradation, which has not been
`demonstrated in car
`radios, may prove to be generally
`imperceptible.
`'3
`
`IV. BLEND WITH TIIVIE DIVERSITY ~
`
`Perhaps the most effective method for dealing with
`the nonstationary mobile radio channel
`is to provide time
`diversity between two independent transmissions of the same
`audio source.
`The FM IBOC DAB concept
`inherently
`provides this ability by delaying the analog transmission by a
`fixed time olfset
`relative to the decoded DAB audio
`
`transmission. When the DAB transmission is blocked (or
`corrupted for any reason) for a short time, then the outage at
`the DAB decoder is heard after the diversity delay; This
`diversity delay is incurred at the receiver and is comprised of
`deinterleaving and FEC ‘decoding delay, audio decoding
`delay, and any additional delay for diversity improvement.
`The FEC decoder can be use to identify faulty audio frames
`and, therefore, the exact time of the DAB audio outage can be
`predicted.
`If the channel becomes unblocked after
`the
`diversity delay, then the analog signal can be demodulated
`such that its detected audio output can be blended in while
`blending out
`the faulty DAB segment. The listener may
`detect the temporary degradation in audio quality during the
`analog blend duration, but will not experience muting or
`undesirable artifacts.
`
`If the diversity delay is sufficiently large siich that
`the DAB and analog outages are independent,
`then the
`probability of an outage atlcr diversity is the square of the
`probability of outage without diversity. For instance, if the
`probability of an outage is
`l.L'I“/o,
`then lite probability of
`outage after diversity is 0.01%, which is a great improvement.
`The actual performance can be quantified with knowledge of
`the antocorrelation function of the channel outage due to
`severe
`impairment.
`- This
`antocoi-relation
`function
`is
`expressed as
`
`R(1'): E{x(r)v xfi‘ — r)}
`
`is defined as the stochastic process; of the
`where X0‘)
`channel loss probability such that a “I” is assigned when the
`
`(cid:51)(cid:68)(cid:74)(cid:72)(cid:3)(cid:19)(cid:19)(cid:19)(cid:19)(cid:27)
`
`
`
`is
`channel is lost and a “O” is assigned when the channel
`clear, and 1' is the diversity delay time offset. The probability
`
`The
`is p = E{x(i)} .
`of outage without diversity
`autocorrelation function represents the probability of channel
`outage after diversity improvement as 21 function of lime
`offset.
`
`Figure 15. Example Autoeorrelation Function of channel
`loss due to blockage or severe impairment (p=l].04).
`
`An example aulocorrelation function is shown in
`15; however, an actual autocorrelation function
`Figure
`depends on distance from the station,
`terrain, propagation
`conditions etc. The figure shows that if the analog signal is
`not delayed relative to the DAB signal (zero time delay), then
`the outages are correlated and no benefit
`is gained from
`blending since the probability of outage remains the same as
`without diversity.
`If the delay is large, then events become
`uncorrelated and the probability approaches the square of the
`probability without diversity.
`The blend feature also solves the problem of fast
`tuning time. Without blend. a receiver would incur the
`diversity delay after tuning to a station before the listener
`hears the audio.
`The blend feature will demodulate the
`
`analog signal almost instantly, allowing the listener to hear
`the selection before blending to DAB several seconds later.
`
`V. ALL-DIGITAL DAB WITH TIME DIVERSITY
`
`The IBOC designs permit evolution to an all-digital
`DAB format. without the host analog signal present. the
`DAB will be transmitted within the primary spectral channel
`allocation (i 100 kHz). Adjacent channel interference issues
`are alleviated. The DAB power can be increased by as much
`as 20 dB. riihrtantisllv increasing the DAB coverage area.
`The
`transmission
`format will
`include
`data.
`normal
`compressed audio, and a more compressed version of the
`same audio signal which is delayed by the diversity time
`
`429
`
`FEC with interleaving is applied to the data and
`offset.
`nonnal compressed audio. The lower rate audio signal
`is
`used for blending during outages in place of the analog signal
`in the IBOC technique. Furthermore this lower rate audio
`signal employs FEC coding without interleaving. Therefore
`the all-digital signal format facilitates fast tuning and exploits
`time and frequency diversity with the lower-rate redundant
`digital audio signal.
`
`V]. CONCLUSIONS
`
`A brief FM Hybrid IBOC DAB system overview was
`presented.
`Careful
`spectral placement and power
`level
`settings assure minimal impact to the coexisting FM signal.
`while modulation and new FEC coding tailored specifically
`for this are designed to provide robust performance.
`The impact on perfonnance of the host FM signal in
`the presence of various IBOC DAB configurations has been
`analyzed. Detailed simulations and analysis indicate that FM
`performance is
`least affected when a pulse-shaped DAB
`signal is placed between 129 kHz and 199 kHz from the FM
`carrier. Modulation and coding tradeoffs have been exercised
`to provide the spectral efficiency required to fit
`the DAB
`signal within this bandwidth.
`.
`The FM IBOC DAB system will provide virtual-CD-
`quality stereo audio using redundant spectral sidebands to
`provide frequency diversity and immunity to first-adjacent
`interference. Time diversity is provided through interleaving.
`A blend-to-analog feature, with time diversity on the order of
`seconds, permits virtually instant tuning time while filling
`DAB audio gaps due to blockages or severe impairments.
`This