SINGLE FREQUENCY NETWORKS IN DIGITAL RADIO Anders Mattsson and John Kean Harris Corp and NPR Labs Mason OH Washington DC

Transcription

1 SINGLE FREQUENCY NETWORKS IN DIGITAL RADIO Anders Mattsson and John Kean Harris Corp and NPR Labs Mason OH Washington DC ABSTRACT Not too surprisingly, a Digital Radio Single Frequeny Network (SFN) shares the same properties as a Digital Television SFN. For this reason some of this paper has been taken from [1]. The main differenes between radio and TV SFN are due to the different S/N requirements and guard intervals. SFNs do offer signifiant potential advantages inluding better overage, less interferene, less power, higher reliability and a more effiient spetrum use. These properties are derived from basi propagation models. Potential problems that must be onsidered, and are speifi to SFNs, are disussed. They all relate bak to reeiver performane. In addition, some basi disussion about how antenna patterns an be used to ombat delay problems is inluded. The hybrid mode of HD Radio has some signifiant differenes in its requirements for SFN distribution sine both the digital and analog omponents must be onsidered. For a digital-only SFN overlaid on the overage area of a hybrid FM station, there would still be analog impliations to onsider. Sine hybrid transmission is likely to be the reality for many years to ome, it raises speial issues on whih we elaborate later in the paper. GENERAL Aording to Shannon the more energy the more information (bits) an be transmitted sine the bit energy versus the noise power spetral density inreases (E b /N). All broadasters know that the more power you transmit the better overage you get. This has always been understood to mean more power into the same antenna. However, the same effet an be had by sending the extra power to another antenna. In a sense, one an argue that the SFN onept is nothing but more power in disguise. There is some truth to that opinion, but it will be shown that for pratial propagation, SFNs should atually over the same area with less power. The reason is that an SFN allows a more even distribution of the power. So, in theory, adding transmitters, i.e. building a Single Frequeny Networks (SFNs) should always improve performane. Single Frequeny Networks (SFNs) are nothing new in radio; they have been around for a long time in analogue FM. However, performane has been less than stellar. This poor performane is not due to any inherent limitation in SFNs. In fat, it is beause of the lak of equalizers in traditional FM reeivers. At the moment the reeiver an handle the multipath, SFNs offer many potential advantages. Sine all digital systems suh as HD Radio and DRM already have equalization, the old limitations are gone.1 This opens the door for SFNs. What SFNs have to offer are: flexible overage, improved overage, dereased interferene, and higher reuse. Sine all SFN systems are inherently the same, in partiular for TV and radio, muh of the theoretial bakground has been taken from [1]. To inrease overage in single transmitter system requires a ombination of inreased antenna height, inreased output power, and/or a different antenna pattern. None of these options might be pratial. In this ase SFNs an offer an attrative option, easily extending overage with the simple addition of lowerpower transmitters at various sites throughout the desired overage area. Among their many benefits, SFNs are more flexible in terms of overage area, superior interferene performane and inherently more fault-resistant. Another differene for a SFN or repeater may be ontrol of interferene to stations on the same or adjaent frequenies sine lower power (than the primary broadast transmitter) to fill areas of poor overage. The fous is on HD Radio and DRM, but the same priniples apply to all SFN systems. Three different ases of digital audio broadasting (DAB) will be onsidered, of whih the first two are very similar: digital-only HD Radio and DRM. The third being hybrid analog-digital version of the HD Radio system. MULTIPATH AND DOPPLER SHIFT No reeiver an distinguish between refleted signals from one transmitter versus several reeived signals from multiple transmitters. To the extent that a system an handle multipath, it is possible to design an SFN around it. Sine analogue AM and FM reeivers do not 1 HD Radio is an In-Band On-Channel (IBOC) DAB tehnology liensed by ibiquity Digital Corporation and approved by the FCC and National Radio Systems Committee. DRM (Digital Radio Mondial) is an opensoure IBOC platform for shortwave, mediumwave/am and longwave digital radio broadasting approved by the ITU, IEC and ETSI.

2 have equalizers, how do they survive in a multipath environment? How do they funtion at all, sine some multipath is always present? The answer lies in the narrow bandwidth in relation to the delay spread. The delay spread being the duration of the RF hannel's impulse response. For most pratial ases, the impulse response will have died down within 100 µs, often after only a few µs. The net effet is that the reeived signal will appear to fade in amplitude. I.e. there is no frequeny seletivity, ommonly referred to as "flat fading." In the ase of analog stereo FM, however, the effets of multipath an our before flat fading is signifiant. For example, delays of only a few miroseonds between the diret and refleted paths an ause audible distortion and rosstalk if the amplitude ratios are small (e.g., less than 0 db). This is beause analog stereo is sensitive to sideband distortions farther from the hannel enter than with monophoni FM modulation, and is analogous to wideband data modulation requiring more equalization against multipath than narrowband data. If the bandwidth of the signal inreases, the fading will beome more and more frequeny dependent. In the time domain, the effet of non-flat fading is inter symbol interferene (ISI), i.e. the bits/symbols start to overlap eah other. It is intuitively lear that a small amount of overlap should be fine, but signifiant amounts will blur the signal. In a digital broadast system with its higher bandwidth and need for high data rates, intersymbol interferene does beome a problem. There are two basi approahes to ombat multipath propagation/isi. One is to design a signal that is robust to reasonable multipath. The seond is to have an equalizer in the reeiver. Quite often a ombination of the two is used, starting with the latter. Sine the RF hannel an be modeled as time a varying filter, the job of the equalizer is to ontinuously find the inverse filter and apply it to the reeived signal. The main problem is generally in the estimation of the filter. This has been, and still is, an area of ative researh. The seond approah is to use a signal that is inherently immune to multipath. With this respet OFDM signals have beome extremely popular. They are used in HD Radio and DRM as well as some digital television systems. An OFDM system with N arriers an be thought of as onsisting of N narrowband transmitters eah transmitting a part of the signal. The resistane to multipath is based on two properties, a guard interval and the use of orthogonal arriers. The atual signals onsist of CW arriers, whose phase and amplitude are kept fixed during the symbol time. To make them orthogonal, the spaing must be a multiple of the inverse of the symbol time. Assuming that the hannel impulse response is shorter than the guard interval, the guard interval between the symbols ensures that the intersymbol interferene period is longer than pratial delay spread. For this reason, many OFDM systems an work in different modes, allowing the user to hoose different guard intervals. The frequeny response aused by the RF hannel will only ause a fixed phase and amplitude offset to eah arrier, resulting in eah arrier seeing flat fading, making them easy to detet. It is worth noting that applying a mathed filter to eah arrier yields an optimum linear detetor, equivalent to taking the FFT on the reeived signal. This gives the OFDM signal an additional advantage of signal proessing with redued omplexity resulting in lower ost reeivers. As with all systems, there are trade-offs; the guard interval results in a dereased throughput. An OFDM system without guard interval using an equalizer would have a higher throughput. It should also be noted that it is possible to add an equalizer to an analogue FM reeiver, given the dereasing ost of proessing power (DSPs and FPGAs), whih would make analogue SFNs muh more tempting. DOPPLER SHIFT Doppler shift will our if the impulse response hanges over time, for example by the reeiver being in a moving ar. The net effet is that a refleted signal an be slightly offset in frequeny f = f 1 1+ v / v v Speed of objet v speed of light 3 10⁸ m/s In an SFN system when moving away from one transmitter towards the other, the frequeny differene between the two signals will be f 1 = f 1+ v / v 1 f 1 v / v f v For a ar on a freeway, worst ase, this would be about 100MHz(( 100km/h)/(3 10⁸ ))=/3 100/3.6=Hz.

3 In an OFDM reeiver without equalizer, the two signals, after taking the FFT, will appear as having a Hz offset. This destroys the orthogonality between the arriers. An alternative view of this degradation is as intersymbol interferene in the frequeny domain. The question is: how muh interferene this will ause and might it ause problems? Assuming that the reeivers do a simple FFT of the signal, then the time limited arriers will have a sin(πx)/πx type of spetra, where eah arrier would be found at x=1,,3... et. For FM HD radio with a 363.4Hz arrier spaing, a Hz offset will result in the n'th arrier leaking into the first by ε = sin(( π / n) /( π ( / n)) n n = ± 1, ±, ± 3,... Sine the different arriers are assumed unorrelated, the power will be ε n SDR = 40dB Hene, Doppler shift is not a problem. If someone would be driving really fast, say 00km/h (15mph), the Doppler shift ould reah 44 Hz, and the SDR would be 14 db. For a QPSK signal, this is still aeptable. Note that this assumes that the ar is driving on a straight line from one transmitter to the other; in reality this is somewhat unlikely. It follows that multiple transmitters in an DAB SFN an have quite a bit of frequeny offset. As a rough estimate, 10 Hz. By omparison, the ATSC digital television SFN standard alls for transmitters to be within 1 Hz of eah other. The reason lies with the reeiver equalizer, whih needs to trak the two arriers, and in many reeivers this happens on a rate of a few Hz. With today's easy aess to good frequeny referenes (GPS), it is relatively easy to lok two transmitters to within one Hz. HD RADIO RECEIVER PERFORMANCE As in the normal multipath ase, the reeiver will work just fine, as long as signals delayed more than what the equalizer an handle will be below the signal to noise threshold - taking some fading margin into onsideration. Note that the transition is smooth, so that if a reeiver an handle 100 us of delay, it won't immediately break down at 101us. Theoretially, what really matters is the bit energy (E b ) relative to the power spetral density of the noise (N0). Sine signal to noise (S/N) is more ommonly measured, it is pratial to relate the two. Sine the HD Radio arriers are QPSK, one an approximate the S/N needed by relating bak to QPSK performane. In turn, QPSK an be seen as two orthogonal BPSK signals, so it is possible to relate bak to BPSK performane. Turns out that Eb/No for an OFDM arrier is the same as the S/N of the signal. Half of the QPSK signal energy and half the noise energy is in the I hannel and the other half in the Q hannel. The bit error beomes ( S N ) P B = Q / For an SNR of 5 db, the BER is already down to 10e- 6, so the signal is inherently robust. The HD Radio waveform has a guard interval of about 156 µs, out of a total length of.9 ms. If there is a multipath delay of 78µs, there will be little overlap within the "ore" symbol, and very little ISI. For a multipath delay of 156 µs, the overlap will be 78 µs, where the overlapping signal is gradually dereasing due to the pulse shaping. Hene the S/N due to the overlap will be 10log 10 (900/78)=16dB. Sine this is a good S/N ratio for QPSK, it seems that multipath delays of up to the guard interval of 156 µs, should be aeptable. Ideally it should be less than 78 µs. Should one allow for a fading margin? Definitely, the different paths will fade independently so a 10 db fading margin should be suffiient. This only affets delayed signal paths that are beyond 78 µs. For a delayed signal of 156 µs, the S/N will be 6dB taking the fading margin into onsideration, whih is an aeptable interferene level. DRM RECEIVER PERFROMANCE The DRM signal has a guard interval of more than ms, as listed in Table 1, or about ten times that of HD Radio. The standard allows for 4-QAM (QPSK), 16- QAM and 64-QAM. The S/N requirement in QPSK mode will be the same as for HD Radio. The other modes will require higher S/N ratios of 6 db and 1 db relative the QPSK mode. Table 1 DRM, different modes As ompared to HD Radio, the guard and symbol time are both about five to ten times greater. The advantage is that this OFDM signal an handle larger delay

4 spread. The down side is that it is more sensitive to Doppler shift. SFN IMPLEMENTATION ISSUES The problems one is likely to enounter are the same as in TV [1]. Implementing a SFN does take are sine there are several potential problems. Simply feeding the same AES data streams to two different transmitters will not work. For an SFN system to work the transmitted signals must be essentially idential and within an appropriate time interval. There are at least three things that an break an SFN system: Timing errors, frequeny offset, and non-idential data. In an SFN system most reeivers will reeive only one dominating transmitter. Dominating in the sense that the other signals are suffiiently weak to not interfere. The only reeivers that will experiene multiple signals are those where the signals from the different transmitters are within the SNR range of the various systems, e.g., db. Outside this range, the other transmitters an be treated as noise, as shown in Figure 1. Signals from other transmitters outside this range an be used by a reeiver with a good equalizer, but it doesn't have to. In an OFDM system, all the signals arriving within the guard interval will be used by the reeiver. These signals will be used by the OFDM reeiver without any need for an equalizer. TX1 TX Figure 1 Areas served by the different transmitters In an SFN system there will always be areas where the signals from two or more transmitters are very lose in amplitude, within a db or less. This will result in some frequenies being nothed out. There is nothing that an be done about this. However, it will not break the system; the error orretion will handle it (within reason). The problem with signals lose in amplitude is less of an issue in OFDM sine OFDM reeivers do not are if the hannel is minimum phase or not. As long as the delayed signal in an OFDM signal are within the guard interval, they will be orrelated and will make the RF hannel appear to have a more uniform frequeny response. A worst ase is two signals of equal amplitude with a delay of µs, whih will result in nothes in the frequeny response every 13-7 khz. For HD Radio s 363 Hz frequeny spaing, there is more or less a 50% hane that a partiular arrier is signifiantly attenuated. FREQUENCY ERRORS It was shown earlier that any frequeny offset between arriers in an OFDM system results in ISI in the frequeny domain. Further more, this frequeny offset an be seen as a Doppler shift. As long as the frequeny offset is within the Doppler shift bound, the system will work. For both HD Radio and DRM this limit is a fairly generous 0Hz (approx). In systems that do use equalizers, the Doppler shift an be traked out providing the equalizer an be made to trak the hannel. In OFDM systems without equalizers, the general ase, it is not possible to trak it out. DATA AND SYNCHRONIZATION ERRORS Ideally, the different transmitters will transmit exatly the same signal. One way to ahieve this is by distributing the atual RF signal using repeaters. If not, the individual transmitters must perform idential modulation. If the individually modulated signals are not the same, the reeiver will obviously not work. This is a problem unique to SFN systems. The studio to transmitter link, STL must be error free; if not, this beomes a separate problem. It might seem that this would be all. However, not all digital transmission standards are deterministi. For example, the input data is often proessed in bloks, but where to start is left up to the modulator. Trellis oders might have random initial states, as in the ATSC digital TV standard. Or, if an error ours in a trellis oder, it might propagate forever. In this ase the system is, in some sense, unstable. Furthermore, any training and synhronization sequenes must be inserted at exatly the same point. These types of problems are also unique to an SFN. One way to ahieve synhronization is to insert some symbol in the data stream. An alternative is to send all, or part of, the modulated data to the transmitter. This simplifies things but might require more bandwidth in the STL. Data synhronization only beomes a problem if SFN appliations were not onsidered at the system design phase. One one is aware of the problems of data

5 synhronization, they are easily irumvented at the time of the standard setting. DELAY AND TIMING ERRORS The theory behind signal delay is exatly the same for both radio and TV SFNs. This and the following setions up until Less Interferene in SFNs summarize results found in [1]. For the SFN to work the time offset as seem by the reeiver must be within the bounds of the equalizer. The time differene between the signals from two different transmitters depends on two fators: the time offset between the transmitters, and the reeiver's position relative to the transmitters. If the delay is longer than what the equalizer an handle, there will be problems. Similarly, if the reeiver already sees a delayed signal from one transmitter, adding a seond or third transmitter et., whih introdues very little extra delay, an potentially put the equalizer over the edge. depends on the distane raised to some power α (propagation onstant). Setting the signal ratio from the two transmitters in Figure 1 onstant and solving gives: P P TX 1 TX = 1+ p x 1 p ( ( x ) + y ) ( ( x + ) + y ) + y = p 4 p ( 1 ). p This is the equation of a irle entered at (((1+p)/(1- p)),0) with radius (( p)/(1-p)). The irular shape does not depend on the attenuation onstant α, but the atual signal ratio on the irle does. The higher the attenuation, α, the higher the signal ratio for a given irle, Figure. α α = α Lines of onstant time differene turn out to be hyperbolas as seen in Figure. Lines of onstant time differene TX distane is ο<α< α=0 ο<α< α= Figure Lines of onstant delay. Given to transmitters at oordinates +/- (see figure ) the lines of onstant delays are: y = a osh( t) x = a TX1 a sinh( t) TX α= τ=/v Figure 3 Cirles showing onstant signal ratio and hyperbolas showing equal delay. The two transmitters at ±. In reality the urves of onstant signal ratio will depend on the atual propagation and antenna pattern and ould be quite different from irles. The important point is that the urves are generally different from the urves of onstant delay. If the HD Radio reeiver an handle 78µs of delay this distane is about 4 km (15 miles), and for 156µs the distane is 48 km (30 miles). For DRM, the same figures are in above 00 km (15 miles). Far away from the transmitters, the signal strength from the two transmitters will be almost the same. If the distane between the two transmitters is suh that the delay is beyond what the reeiver an handle one will get a dead zone behind the two transmitters, Figure 4. where a is the distane differene, whih diretly relates to the time differene through τ = a / v. In an SFN the delays are only important in relation to the signal strength ratio. This an be alulated assuming omni antennas and that the signal attenuation

6 (α=) they are equivalent, for an α greater than, the SFN will need less power [1]. The greater α is the more advantageous an SFN beomes. For an SFN with N transmitters the ratio will be P sin gle P SFN = N α / 1. Note that if α> the overall power onsumption in an SFN will derease as the number of transmitters inrease. Figure 4 The shaded region shows the area where no reeption is possible in an SFN with two transmitters, if the distane between the two is too long. Depending on the systems, this area, assuming that it exists, might be so far away that it is outside the desired overage area. In whih ase it doesn t matter. The antenna pattern an be used to ombat this problem, whih will be overed later. LESS INTERFERENCE IN SFNS This is an important property of an SFN. The main reason for this is that the ratio between the losest transmitter and the interferer is greater in an SFN. The alulations are a bit more involved. The interested reader is referred to [1]. In essene, the more transmitters used in the SFN and the higher that the propagation onstant (α) is, the more advantageous an SFN beomes. TRANSMITTER SPACING This is probably the most important issue and depends on the multipath properties of the OFDM signal; and, if present, the abilities of the equalizer. The previous setion provides a lower bound on the transmitters spaing. It is obviously desirable to find the maximum spaing. Taking into aount that the system an handle some harmful interferene (i.e., long delays beyond a ertain level) as long as it is below x db, a less onservative estimate an be found: d x /( 10α ) = = yµ s 300 ( ) m x / 10α 10 1 where y µs is the allowable delay in µs, α is the signal attenuation and x is the aeptable interferene level in db. Using x=15db, α=3, and y=78µs, then =4km (14 miles). So the maximum distane between the transmitters is 48km (8 miles), about twie the distane predited by using y µs 300. For a more detailed derivation see [1]. TX ANTENNA PATTERNS The antenna patterns an be used to minimize areas of harmful multipath. If one of two idential transmitters hanges its output power, then the area between the two where the signals are within ±x db will derease, and move towards the transmitter with the lower power. As an example: assume two transmitters have the same output power and omni antennas, let the distane between them be d= and let α=3. Then the region where the signals are within 15 db of eah other ours at distane of from eah transmitter. In this ase, then the signals are within 15 db in the interval -0.5 to 0.5, or about 50% of the distane. If the signal from the seond transmitter is 30 db weaker, this interval beomes 0.5 to 0.95 or 5% of the distane, Figure 5. POWER CONSUMTION IN SFNS It turns out [1] that the power onsumption in an SFN relative to a single transmitter system depends entirely on the signal attenuation α. For free spae propagation

7 to use four transmitters on a irle with diretional antennas and a fifth transmitter in the enter using an omni diretional antenna. In this ase, the antenna gain in the diretional antennas must be suffiiently low outside ±90deg in addition to a good front to bak ratio. Other onfigurations are obvious. Figure 5 The ratio S 1 /S for α=1 with one transmitter at -1 and one at +1. The same effet an be had if the transmitters that are away from the enter use a diretional antenna, where the front to bak ratio of the antenna provides the power differene. This will derease the area of harmful interferene, i.e., signals outside the guard interval/equalizer range, allowing wider transmitter spaing. Note that the delay between the transmitters must be hanged so that zero delay between the reeived signals ours in areas where the two signals are of equal magnitude. ) Figure 6 The effet on the area where equalization is needed, thik line, for omni and diretional antennas. If diretional antennas are used, as shown in Figure 6, it is probably better to delay the signals in the two outlying transmitters by the propagation delay. This approah may eliminate the undersored "dead zone" by using an antenna with a front-to-bak ratio equal to the S/N ratio plus a suitable fading margin (15 db for HD Radio), Figure 7, but this omes at the prie of lower signal strength in the area between the antennas. Adding a third transmitter between the two as shown in Figure 6 an solve this problem. Another alternative is ( S /S 1 >15 db Figure 7 Using diretional antenna with suffiiently high front to bak ratio to eliminate the "dead zone." OTHER POTENTIAL ADVANTAGES S 1 /S >15 db There are two potential advantages of interest: S/N ratio improvement and diversity. The first one is based on the fat that the sum of two idential signals with independent noise results in a 3 db S/N ratio gain. In an OFDM system, any signal arriving within the guard interval will improve the signal to noise ratio and is in theory helpful. In a system using equalizers, the S/N ratio gain might not happen due to the equalizer itself adding more gain to parts of the spetrum where the signal is weak. This tends to result in a noise gain. The seond advantage is that multiple signals provide diversity. If the reeiver sees two signals that are fading independently, the probability that both signals are drowned by the noise is signifiantly lower than for a single signal. For example, if the probability that one signal is below the noise level is 5%; the probability that both are below the noise level is only 0.5%. However, this assumes the two signals to be independent, whih is not the ase for OFDM signals when the delay is shorter than the guard interval. Antenna diversity at the reeiver would help, but is only realisti for some reeiving situations, suh as ar radios. A fixed radio reeiver would benefit from a diretional antenna. Suh an arrangement would allow the reeiver to hange the ratio of the two signals. For a portable reeiver, this is unrealisti.

8 PLANNING AND IMPLEMENTATION An hallenge to designing SFNs is ontrolling the interferene between the transmitters. One must predit the signal strength from eah transmitter as well as the relative propagation delay times aross the desired overage area. When signal strengths are similar and differential delay times exeed a ritial value, multipath interferene is predited. The objetive is to minimize these areas of multipath or move them to relatively unpopulated plaes. In designing SFNs, multipath an be ontrolled by transmitter site seletion, antenna radiation pattern, output power, and modulation time offsets. The site seletion is the most powerful, so areful planning of the system is important. Changing sites one the system is built may be expensive or prohibited. Optimizing the antenna radiation pattern and signal delay are possible one the sites are established. Output power should be established one the site is hosen sine a 5-10 db power inrease usually requires extensive re-engineering of the other transmission parameters. A power derease results in a waste of installed transmitter apaity. No system, SFN or not, will provide 100% overage. As with all broadast systems, there will be a tradeoff between overage and ost. The multipath requirement is the same as for a single transmitter system (i.e., it depends on the system and to some extent what margins are used). broadast system that an not easily be synhronized due to non-deterministi modulation, repeaters an be used to reate an SFN. Repeaters have at least one short oming: their interonnetion link adds delay. This is a serious issue sine it is often desirable to have zero delay right between the two transmitters. One an employ synhronized transport networks and GPS frequeny standards, but the osts may outweigh the overage benefits offered by the repeater. Figure 9 KCSN(FM) primary analog overage area (part); green is dbu. This Longley-Rie overage predition shows a sharp utoff of overage to the south. Central Los Angeles is in the lower right of the map. The planning of repeaters, boosters and single frequeny networks is greatly aided by omputer analysis. Computer models an evaluate the field strength ratios and propagation time differentials of the transmitters at millions of loal points, a proess that would be nearly impossible by hand. Interative analysis of the loation and severity of multipath zones permits the engineer to hoose transmission sites and other parameters neessary to optimize a SFN design. Figure 8 Top: Site loation that will inrease multipath. Bottom: Site loation that will derease multipath. In the SFN site seletion proess the trik is to use the terrain to one s advantage. Figure 8 shows two ways to ahieve overage in a hilly area, with most of the population living in the valleys. With the transmitters plaed on the top of the ridges, the risk of exessive multipath is signifiant. Moving the transmitters down into the valleys eliminates most of the multipath. Repeaters (also known as boosters) an be used to implement SFNs and they do have one advantage: sine the repeater merely repeats the data modulation, the transmitted RF signals will be idential. Even for a Figure 10 Combined KCSN primary and booster overage with F(50,50) 60 dbu ontours overlaying Longley-Rie field strength predition.

9 As a ase study of SFN design, NPR Labs wanted to evaluate the performane of the first HD Radio single frequeny network, built by publi radio station KCSN(FM), Northridge, California [5]. This station experienes terrain shadowing effets, aused by the Santa Monia Mountain range that extends along an east-west line in the southern part of their overage area. This effetively shields Santa Monia, Beverly Hills and Hollywood from servie, as shown in Figure 9. KCSN designed and built a hybrid (analog FM and HD Radio) booster on a building in south Beverly Hills with a diretional antenna array aimed northward. This filled in signal in the shielded area, but avoided overage extension beyond the 60 dbu servie ontour. The predited overage with the booster added is shown Figure 10. The KCSN booster and primary transmitter are fed by time-synhronized STLs, so that digital audio is delivered to the inputs of the analog and HD Radio transmitters with differential delay of ± µs. GPS is used at both transmitters to ontrol operating frequenies. To evaluate multipath intersymbol interferene NPR developed omputer software to model KCSN s twotransmitter SFN. First, the terrain-sensitive overage preditions were performed in a RF design tool using the Longley-Rie propagation model with 3-ar seond resolution USGS terrain data. A reeive height of meters was hosen to represent ground-based (espeially vehiular) overage. Land-use land-over adjustments were used to improve auray of predited fields. Next, numeri arrays ontaining the propagation study for eah transmitter were imported into MapInfo, a GIS tool, where ustom software was developed to: Calulate signal propagation time from the primary and booster transmitters to a grid of finely-spaed points aross the study area; Compare the field strength ratio of primary and booster signals at the same points as above; Determine field ratios and time-of-arrival differenes that may result in intersymbol interferene of the HD Radio signals from primary and booster transmitters); and Generate a map showing the loations that exeed the parameter guidelines. Figure 11 Map showing loations of potentially high primary and booster multipath as olored dots. Reddish dots indiate where the signal from the primary transmitter is stronger and bluish dots indiate where the signal from the booster is stronger; the signal ratios of the gray dots are within 1.5 db. The retangular box shows the study area of the measurement test map shown below.

10 Figure 11 shows the result of the multipath model as a geographi map, where loations of potentially high multipath are predited. This depition is based on field ratios within 10 db and signal propagation differene of greater than 75 µs, as disussed earlier for HD Radio. To eliminate loations that are below pratial reeption, only field strengths greater than 50 dbu for the stronger signal are shown. The loations are olor-oded to indiate whether the primary or booster signal is stronger, although this is unimportant from a multipath standpoint. It is apparent that the area near eah transmitter, where its signal dominates, is free of multipath. A zone of potential multipath rings eah transmitter at greater distanes, depending on the radiated power, distane and terrain attenuation effets. In the booster s ase, the signals mix along the south ridge of the mountains (north of Santa Monia and Beverly Hills) and south of the booster (near Marina Del Rey, Culver City and Los Angeles, whih are mostly in the signal fringe). To evaluate the multipath predition model NPR Labs used its HD Radio Logger to ollet digital reeive status, analog field strength, GPS loation and time. Figure 1 shows an enlarged portion of the KCSN overage surrounding the transmitter. The measurement van s drive-test route is shown as a series of small boxes, in whih the perentage of loal digital reeption (as a funtion of measurement time) is blak for % availability, gray for 90-97% availability and white for less than 90% availability. Most areas experiene high availability of digital reeption (blak squares). The areas of adequate field strength (blue, green or yellow shading) with low availability (white squares) suggest onditions where intersymbol interferene may degrade digital reeption. In Figure 1, the areas of low availability appear in East Los Angeles on US Hwy. 101, along Mulholland Drive, north of Beverly Hills and through the anyons of the Santa Monia Mountains on I-405. Comparison with the Figure 11 shows good agreement with these predited multipath areas. This supports the need to use a multipath model in the design of single frequeny networks, to determine the extent and loation of intersymbol interferene. Multipath models work well for FM reeption using suitable parameters. However, analog stereo has far less tolerane to differential delay than digital systems suh as as HD Radio. Multipath delays of greater than 1- µs will ause audible distortion and rosstalk. The intolerane to multipath begins at field ratios of approximately 0 db. Consequently, single frequeny Figure 1 Measurement of HD Radio reeption around the KCSN booster. A Longley-Rie field strength predition underlay shows loations where the booster's field strength is dbu (blue) or greater.

11 networks are likely to reate large areas of multipath effet with analog FM stereo. Most suessful analog SFNs have substantial terrain shielding between the primary and booster transmitters, or the zones of multipath an be shifted to sparsely populated areas. HYBRID SYSTEMS In a system suh as hybrid HD Radio using both digital and analogue arriers, the whole question about SFNs beomes ompliated. As disussed above, the digital part of the signal is not a problem, but the analogue part will suffer from the known problems of analogue SFNs. (Note that the problem is not a fundamental imitation with analogue signal; the problem is that the analogue signal, just as many digital signals, requires an equalizer to work in a multipath environment.) A reeiver that an handle digital signals does have the proessing power to implement an equalizer for the analogue signal. But sine existing analogue reeivers lak one, it is presently a moot point. CONCLUSIONS Digital radio lends itself naturally to SFN implementations. The main advantage is the potential of very flexible overage and easy expansion - simply add more transmitters. Depending on the length of the guard interval is, some are will be needed to avoid exessive multipath. The hybrid IBOC system warrants some further studies with respet to the analog part of the signal, before it is lear how well SFN will work in this ase. For all other systems, SFN should not pose any fundamental RF related problem. REFERENCES [1] Anders Mattsson, "Single Frequeny Networks in DTV," 53rd annual IEEE broadast symposium pro, Washington 003. [] Ibiquity, "TX_SSS_1038 Common Amplifiation," Ibiquity [3] EBU-UER, "ETSI ES Digital Radio Mondial (DRM); System Speifiation," 005. [4] John Kean, HD Radio Coverage Measurement and Predition 006 NAB Engineering Conferene Proedings.