Pulished IEEE Transactions on Broadcasting, vol. 5, no., 005, pp. 3-9 A Hierarchical Modulation for Upgrading Digital Broadcast Systems Hong Jiang and Paul Wilford Bell Laoratories, Lucent Technologies Inc. 700 Mountain Ave, P.O. Box 636, Murray Hill, NJ 07974-0636 Astract A hierarchical modulation scheme is proposed to upgrade an existing digital roadcast system, such as satellite TV, or satellite radio, y adding more data in its transmission. The hierarchical modulation consists of a asic constellation, which is the same as in the original system, and a secondary constellation, which carries the additional data for the upgraded system. The upgraded system with the hierarchical modulation is ackward compatile in the sense that receivers that have een deployed in the original system can continue receiving data in the asic constellation. New receivers can e designed to receive data carried in the secondary constellation, as well as those in the asic constellation. Analysis will e performed to show the tradeoff etween it rate of the data in secondary constellation and the penalty to the performance of receiving the asic constellation. Index Terms ackward compatiility, it rate, constellation mapping, digital roadcast, hierarchical modulation, iterative decoding, local repeaters, penalty analysis, QAM, QPSK, receiver design, system upgrade I. INTRODUCTION Digital roadcast systems have increasingly een deployed for services such as terrestrial digital TV, digital radio, satellite TV and radio, and digital cale TV. A digital roadcast system utilizes regulated frequency ands with fixed andwidths. The capacity of a digital roadcast system is limited y transmission power of the system and channel impairments. Since in a roadcast system, the same data is transmitted to all users, there is a tradeoff etween transmitted it rates and intended coverage areas. A digital roadcast system is usually designed with a it rate that can e relialy received y users in an intended coverage area for a given transmission power. Some digital roadcast systems are designed with a flexiility of transmitting with different it rates y allowing different modulating constellations (such as QPSK, 8PSK, 6QAM, 64QAM etc), and error correction codes of different coding rates. It is desirale for receivers in such a system to e capale of receiving data at all of the specified it rates. An advantage of such a system is that it is easy, at least in concept, to upgrade a deployed system to increase the it rate, simply y switching to a modulation with a larger constellation and/or an error correction code of higher rate. A disadvantage of such an arrangement is that all receivers have to e designed with capaility of receiving all specified rates, and hence such receivers may e expensive to manufacture. Many deployed systems, however, do not have such flexiility; they are designed with a fixed modulation scheme, and error correction codes of fixed rates. Examples of such systems are the original ATSC terrestrial 8VSB system, and some digital satellite radio and TV systems. Oftentimes, after such a system is deployed and in service for a period of time, there are rising needs to upgrade the system to provide more services than the system was originally designed. The needs for upgrading the system may arise for many reasons. There is a market pressure for a service provider to add more services in order to differentiate itself from its competitors. New services may also generate additional revenue. On the technical side, new algorithms may have een developed to allow higher it rate to e transmitted through the same channel, and at the same time, advances in technology make the implementation of the new algorithms feasile and cost-effective. This advance in algorithm development and technology makes it possile to uild a new generation of receivers that can have much etter performance under the same channel conditions. For example, the utilization of iterative decoding algorithms such as BCJR [,, 3] and the MIMO technology [4] can provide additional gains over receivers using traditional algorithms, or a single antenna. Backward compatiility is a major difficulty in increasing the it rate of a fixed rate digital roadcast system that has een already deployed for a period of time. Since the system has already een deployed, there
Coded secondary information its Coded asic information its QPSK Mapper (distance d ) I Q Secondary Mapper (distance d ) I' Q' Modulator x(t) Figure QPSK/6QAM Modulation are a large numer of receivers already used y suscriers that can only receive signals from the original system with the specific modulation and it rate. To replace these deployed receivers y new generation of receivers may e prohiitively expensive. Therefore, any upgrade of the system must e ackward compatile, in a way that the deployed receivers must continue to operate in the upgraded system, even though the deployed receivers may not e ale to receive the supplementary data that is added to the original system. In this paper, we propose a hierarchical modulation [5, 6, 0] method to increase the it rate of a deployed digital roadcast system. The upgraded system is ackward compatile, so that the upgrade is transparent to the deployed receivers of the original system. The hierarchical modulation has already een included in the DVB-T standard []. We will use the term asic information to mean data transmitted in the originally designed system, as is customary in the literature of hierarchical modulations. The data that is added in the upgraded system will e named secondary information. Thus, the originally designed system with deployed receivers transmits asic information only, while the upgraded system transmits oth asic and secondary information. Although the concept of this paper may e applied to hierarchical modulation of any constellations, we will limit our discussions to QPSK/6QAM hierarchical modulation in the remainder of this paper. The reason is twofold. First, most digital satellite systems use QPSK, and due to its small constellation size, a QPSK system not only has the most urgent need for, ut also can enefit the most from the addition of a secondary information channel. Secondly, the use of a specific hierarchical constellation, QPSK/6QAM in this case, simplifies the analysis. The remainder of this paper is organized as follows. In the next section, the hierarchical modulation for the asic and secondary information is descried. An analysis of the BER in terms of CNR and the ratio of minimum distances is given in Section III. Also, in Section III, penalties of the hierarchical modulation to the originally designed receivers are derived. In Section IV, a design of new generation receivers is presented for the hierarchical modulation in which an iterative decoding algorithm is used to provide additional channel gains for oth asic and secondary information its. The it rate at which the secondary information can e relialy received is also estimated. Finally in Section V, we address the issue of the local information transmission. II. QPSK/6QAM HIERARCHICAL MODULATION Hierarchical modulations were initially proposed to provide different classes of data to users with different reception conditions [5, 6, 0]. In that application, all users within the coverage range can relialy receive asic information, while users under more favorale conditions can receive additional refinement information. In the context of the current application of upgrading a deployed QPSK system, the hierarchical QPSK/6QAM modulation can e descried y the lock diagram in Figure. The asic and secondary information its are channel encoded. The coded asic information its are mapped to the QPSK constellation in the same way as the original QPSK system. The minimum distance etween points in the QPSK constellation is denoted y d. This represents the asic hierarchical constellation. The asic hierarchical constellation is next modified according to the coded secondary information its, and the comined hierarchical constellation is formed as shown in Figure (). The comined constellation is a 6QAM constellation with the minimum distance etween two points denoted y d. The mapping of the asic and secondary information its indicated in Figure () is a Karnaugh map style Gray mapping. In Figure (), the
two most significant its (in old face) represent the asic information its while the two remaining its represent the secondary information its. 000 00 0 Figure QPSK/6QAM Hierarchical Constellation (a) QPSK constellation () 6QAM constellation. The its in old face are asic information its, and the rest are secondary information its The two locks, QPSK Mapper and Secondary Mapper, in Figure may e comined into one. They are presented separately ecause the two locks clearly demonstrate the concept of the hierarchical modulation used for upgrading an already deployed system: the secondary information its are added to the transmission y modifying the constellation of the asic information. Let λ e the ratio of the minimum distances in the QPSK and 6QAM, d λ =. () d Q (a) 00 0 Q' 00 000 0000 0 0 00 d d 00 00 0 () I 000 0 00 000 I' λ is an important parameter to characterize the system. If λ = 0, it is the QPSK system, transmitting only the asic information. If λ =, it is the uniform 6QAM, in which the asic and secondary information its have same channel conditions. We will e only interested in the case with 0 < λ <, where the constellation is truly hierarchical. When λ is small, the four points in each quadrant of the constellation form a cloud. To the originally designed receivers, a cloud represents a point in the QPSK constellation. For example, any point in the first quadrant, (the ones laeled 0000, 000, 00, and 000 in Figure ()) is treated as the point 00 in the QPSK constellation. The variation among the points in a cloud has the same effect of white noise on the originally designed receivers. Therefore, in the upgraded system after the secondary information is added, the originally designed receivers will continue to operate and receive the asic information its, with the only difference eing that they may operate at a higher noise level. The additional noise due to the secondary information in the upgraded system imposes a penalty on the performance of originally designed receivers. We will analyze the penalty in terms of the hierarchy parameter λ in the next section. New receivers can e designed to operate in the hierarchical system. New receivers will e ale to distinguish points in the cloud, and extract oth the asic and secondary information its. We will address the issue of new receiver designs in section 4. For convenience, we define the following terminology: QPSK systems systems with QPSK modulation. They refer to the systems efore upgraded to the hierarchical modulation. QPSK receivers the originally designed receivers that are only capale of receiving the QPSK modulation or the asic information in the hierarchical modulation. Hierarchical systems systems with hierarchical modulation, referring to the upgraded systems with oth asic and secondary information. Hierarchical receivers the new receivers that are designed to operate in the QPSK/6QAM system, and are capale of receiving oth the asic and secondary information its. III. PENALTY ANALYSIS 3
The carrier to noise ratio (CNR) of the hierarchical constellation of Figure () is given y Es ( + λ ) d CNR = =, () N 0 N 0 where Es is the carrier power and N 0 is the channel noise power. When signals with the hierarchical constellation of Figure () are received y the QPSK receivers, the constellation is treated as QPSK constellation, with power d. To these receivers, the noise consists of two terms, the channel noise N 0, and the scattering of points in the secondary hierarchy constellation, λ d. We define the modulation noise ratio (MNR) to e the ratio of the power of QPSK constellation to the comined noise power, and it is given y d MNR = N 0 + λ d (3) = CNR. + λ ( + CNR) MNR can e considered as the CNR of the hierarchical constellation of Figure () as viewed y the QPSK receivers. It is a measure of the cleanness of the QPSK constellation as viewed y the QPSK receivers. The performance of these receivers, such as timing recovery and carrier recovery, may e characterized y MNR. Since the performance of these receivers is characterized y CNR in the QPSK system, the performances of these receivers efore and after the secondary information is added may e evaluated y comparing the values of CNR and MNR. The difference etween CNR and MNR is the penalty to the QPSK receivers. From equation (3), the penalty in MNR can e defined as P MNR = + λ ( + CNR), (4) CNR which is the ratio. MNR The penalty, a function of oth λ and CNR, represents the additional carrier power that is needed in the hierarchical system so that the QPSK receivers can see the same cleanness of the constellation as in the QPSK system. It is a measure of how much the QPSK receivers suffer due to the addition of the secondary information. The larger the penalty is, the worse these receivers will perform in the hierarchical system. A plot of P MNR vs. CNR for different values of λ is shown in Figure 3. In most deployed satellite roadcast systems such as satellite radio, or satellite TV, the minimum operating CNR is elow 7dB. That is, receivers can perform relialy at CNR higher than 7dB. As shown in Figure 3, at CNR = 7dB, the penalty is around 0.5dB for λ = 0., and PMNR 0. 5 db for λ = 0. 5. This means that for λ = 0., when the transmission power in the hierarchical system has a 7dB CNR, the QPSK receivers effectively get a QPSK constellation with equivalence of CNR = 6.75dB, ecause the penalty is 0.5dB. For λ = 0. 5, the QPSK receivers can receive a QPSK constellation equivalent to CNR = 6.5dB when the hierarchical constellation has CNR = 7dB. The penalty is even smaller at lower CNR. Although the penalty is higher at higher CNR, the Pmnr (db) 3.5.5 0.5 lamda=0.05 lamda=0.0 lamda=0.5 lamda=0.0 0 5 7 9 3 CNR (db) Figure 3 Penalty in MNR penalty ecomes insignificant ecause there are enough margins in the CNR to meet the desired performance. These results show that the penalty to the QPSK receivers is acceptale for λ 0. 5 under most operating conditions. While MNR may e useful in evaluating performance of timing recovery and carrier recovery, it is not accurate for it error rate (BER) estimate. In the following, we will use another approach to estimate BER of the QPSK receivers in the hierarchical system over additive white Gaussian noise (AWGN) channels. BER computations of QPSK/6QAM over AWGN channels have een performed y Vitthaladevuni et al. [7]. The proaility of raw it error (without error correction coding) made y the QPSK receivers in the hierarchical system is given y 4
λ BER B = Q CNR + λ (5) + λ + Q CNR, + λ x where Q ( x) = erfc, and function erfc(x) is the complementary error function. In the QPSK system efore the secondary information its are added, the QPSK constellation is transmitted and the proaility of it error is BER QPSK = Q( CNR ). (6) A comparison of equation (5) with equation (6) reveals that, for a given CNR, the addition of the secondary information its in the hierarchical system causes the QPSK receivers to have a larger BER, and hence introduces a penalty to these receivers. This penalty in BER is denoted as P BER, and is defined y the following equation CNR λ Q = Q CNR PBER + λ (7) + λ + Q CNR. + λ The BER penalty P BER, also a function of λ and CNR, represents the additional carrier power that is needed in the hierarchical system so that the QPSK receivers can have the same BER as in QPSK system without the secondary information. More precisely, in the hierarchical system with the carrier to noise ratio of CNR, the QPSK receivers can achieve the same raw BER as in the QPSK system with a carrier to noise ratio CNR of. Plots of P BER for various values of λ and P BER CNR are shown in Figure 4. As shown in Figure 4, at raw BER of.e- (corresponding to CNR=7dB), the penalty in BER is less than 0.5dB for λ = 0. and less than 0.5dB for λ = 0. 5. In other words, for the QPSK receivers to achieve a raw BER of.e-, a 0.5dB higher signal power is needed in the hierarchical system than the QPSK system if λ = 0., and a 0.5dB higher signal power is needed if λ = 0. 5. The penalty is even smaller at higher BER. These results are consistent with the analysis of MNR penalty, and they again confirm that the penalty is acceptale for λ 0. 5. IV. DESIGN OF HIERARCHICAL RECEIVERS We now turn to address how to design new generation receivers to process and receive oth asic and secondary information in hierarchical constellation. New generation receivers must e ale to extract all its from the 6QAM constellation of Figure (). Since the asic and secondary information its are channel encoded differently (see elow), they need to e decoded separately y two decoders, and the two decoders can e used to aid each other. A lock diagram of a receiver is shown in Figure 5. After front-end demodulation, the recovered signal is sent to two separate decoders, one for the asic information its, and the other for secondary information its. Both decoders employ some Soft-In- Soft-Out (SISO) decoding algorithms, such as the BCJR algorithm [, 3]. The output of the decoders, usually in the form of proaility or likelihood for a it to e a 0 or, is used in the decision locks to extract the asic and information its, respectively..8 Per (db).6.4. 0.8 0.6 lamda=0.05 lamda=0.0 lamda=0.5 lamda=0.0 0.4 0. 0.E-0.E-0.E-03.E-04.E-05.E-06 Raw BER Figure 4 Penalty in BER 5
Basic Info SISO Decoder Decision Basic Info Bits Recovered Signal Secondary Info SISO Decoder Decision Secondary Info Bits Figure 5 New Generation Block Diagram The channel coding for the asic information is already specified in the original QPSK system, and therefore, the decoder for the asic information its must match the encoder of the original system ecause, one is not free to select a code for the asic information at the time of designing the hierarchical system. Commonly used error correction codes in deployed QPSK systems are RS-convolutional concatenated codes. Although most receivers in such systems use a Viteri-RS decoding, which has hard-decision output, SISO type decoding algorithms exist for RSconvolutional concatenated codes, see [, ]. Unlike the asic information channel, one is free to choose a channel coding for the secondary information its at the time when the hierarchical system is designed. Turo codes [8] or other advanced codes with high coding gain may e used for the secondary information its. These codes are well suited for an SISO type of decoding. We will not dwell on what specific decoding algorithms to use in the hierarchical receivers, ecause they are out of scope of this paper. Instead, we assume that an SISO decoder with a posteriori proaility (APP) decoding [, 3, 8] is used for each of the asic and secondary information decoders, and we propose an iterative method etween the two decoders to improve the performance of oth. In an APP decoding, a posteriori proaility P ( x = y), or P ( x = 0 y), is estimated, where x is the transmitted it, and y is the received symol. A posteriori proaility is the proaility that a transmitted it x is (or 0), given the received symol y. For details of APP decoding, we refer the reader to [, 3, 8] and the references listed therein. To perform the APP decoding, the following proailities need to e computed: ) The proaility that the received symol is y when the transmitted it is 0 and, respectively, P ( y x = ) and P ( y x = 0) ; (8) ) The proaility that the transmitted it is (or 0) x = ) or P ( x = 0). (9) In the following, we will present how to compute these values for the hierarchical modulation of Figure (). The mapping in Figure () allows the real and imaginary components, I and Q, of the received signal to e treated separately. Therefore, when considering proaility values in equations (8) and (9), we consider the real component, I, only, as the imaginary part can e treated identically. The constellation for the real component is shown in Figure 6. In Figure 6, a? represents a it carried y the imaginary component and it is irrelevant in the current discussion.?0??? 0?? 0?0? d d Figure 6 One dimentional hierarchical constellation Let x e the transmitted asic information it from real component, and x s e the transmitted secondary information it from real component. We will compute the proaility values in equations (8) and (9) for each of x and x s. The proaility that the received symol is y when the transmitted asic information it x is is given y More precise notation should e I I x and x s, in order to differentiate these with its from the imaginary component. However, the superscript I is omitted since we are only considering the real component, and there is no risk of confusion. 6
y x = ) = xs = ) y x =, xs = ) (0) + xs = 0) y x =, xs = 0). Similarly, the proaility that the received symol is y when the transmitted secondary information it x s is is given y y xs = ) = x = ) y x =, xs = ) () + x = 0) y x = 0, xs = ). The conditional proaility values in equations (0) and () are readily computed y using the proaility density function of Gaussian distriution, x f ( x) = exp, () σ π σ where σ is the noise power. It can e shown, y x =, xs = ) = f ( y + ( λ) d), (3) ecause x =, xs = corresponds to the point laeled?? in Figure 6, and the point is located at ( d d ) = ( λ) d. Similarly, y x =, x = 0) = f ( y + ( + λ) d ), y x y x s = 0, x = 0, x s s = ) = = 0) = f ( y ( λ) d f ( y ( + λ) d ), ). (4) Now, the proaility values P ( = ), P ( = 0), P ( x s = ), and x s = 0) of equations (0) and () need to e computed. It is in the computation of these values that the iteration etween the asic and secondary information decoders can significantly improve the performance of each decoder. The iteration etween the two decoders can e carried out in the following way: ) Initially, the decoders have no other knowledge on what is transmitted, and hence oth asic and secondary information decoders start with P ( xs = ) = xs = 0) =, (5) P ( x = ) = x = 0) =. (6) By using equations (0), () and (5) and (6), oth decoders proceed to perform APP decoding, and at the end, the asic information decoder computes the (0) (0) proaility values P ( x = ) and P ( x = 0). Similarly, the secondary information decoder computes (0) (0) the proaility values P s ( x s = ) and P s ( x s = 0). ) At iteration k, k=,,3,, the asic information decoder performs APP decoding y using x x ( k ) xs = Ps ( xs t =, 0. (7) ( k ) x = P ( x and equation (0). At the end of decoding, the decoder ( k ) ( k ) computes the values P ( x = ) and P ( x = 0). The secondary information decoder performs APP decoding y using ( k ) x = P ( x t =, 0. (8) ( k ) xs = Ps ( xs and equation (). At the end of decoding, the decoder ( k ) ( k ) computes the values Ps ( xs = ) and P s ( xs = 0). The iteration continues until a preset maximum numer of iterations, K, is reached. The values ( K ) ( K ) ( K ) P ( x = ), P ( x = 0), and P s ( xs = ) and ( K ) Ps ( xs = 0) are then used in the decision locks to determine the asic and secondary information its. It is worthwhile to note that there is a ias in the Karnaugh map style Gray mapping of Figure (). As Figure 6 and equations (3) and (4) demonstrate, it is more likely for a asic information it to e an error it when the secondary information it is than when the secondary information it is 0. For example, it is more likely for the point laeled?? to e mistaken as 0?? or 0?0? than the point laeled?0?. More precisely, the BER of asic information its when the secondary information it is is given y, λ BERB x s = = Q CNR + λ and the BER of asic information its when the secondary information it is 0 is given y 0. + λ BERB x s = = Q CNR + λ Therefore, the transmission of it in the secondary information is unfavorale to the asic information its. If such a ias is undesirale, a mapping in Figure 7 can e used as an alternative. In the mapping of Figure 7, an in the I component of the secondary information its is more likely to cause an error in the asic information, ut a 0 in the Q component of secondary information its is more likely to cause an error. Since I, Q channels are independent and identical, 0 and in the secondary information make the same contriution to the degrading of the asic information its. 7
00 000 Q' 0 00 000 00 00 0 0 d d Figure 7 Balanced hierarchical constellation Next, we analyze the capacity of the channel that carries the secondary information its. It can e shown that the error proaility for the secondary information its over the AWGN channel is λ λ BER S = Q CNR + Q CNR + λ + λ + λ Q CNR. (9) + λ For small values of λ, as is needed to have a small penalty to the QPSK receivers, BER of the secondary information is much higher than that of the asic information. For the secondary information to e received with the same reliaility as the asic information, the channel coding for the secondary information must have a very high coding gain, and consequently, the effective it rate of the secondary information will e much lower than that of the asic information. We now proceed to give a heuristic estimate of the effective it rate of the secondary information its y using a technique of spread spectrum. An estimate of the proaility of errors for a QAM constellation is given in [9, p.338] as P K Q( BT η A CNR ). (0) In equation (0), K is some constant, B is the channel andwidth, and T is the symol duration and d min η A =, () σ 4 A 000 0000 00 000 0 00 I' where d min is the minimum distance etween the points of the constellation, and σ A = ( + λ ) d. For the asic information, T = T, d min = ( λ) d. () For the secondary information, T = Ts, d min = λ d. (3) Sustituting () and (3) into (0), we have the following estimates for the proailities of errors for the asic and secondary information ( λ) P K Q BT CNR (4) + λ λ P s K Q BTs CNR (5) + λ For the secondary information its to e received as relialy as the asic information its, we equate the right hand sides of equations (4) and (5) to get ( λ) Ts = T. (6) λ Since oth asic and secondary information symols carry the same numer of its (two its per symol), equation (6) gives the following relation etween the effective it rates of the asic and secondary information: λ itrate s = itrate. (7) ( λ) This means that for the asic and secondary information its to e received with same BER, the it rate of the secondary information, itrate s, needs to e a factor of λ of the asic information rate, itrate. ( λ) From this heuristic analysis, the secondary information can e relialy received at a rate that is.3% of the transmission rate of the asic information if λ = 0.. The rate of the secondary information can e increased to 3.% of the asic information rate if λ = 0.5. For example, for an existing QPSK satellite radio system with 00 channels of music, the hierarchical system can transmit additional data in equivalence of. channels, for a penalty of 0.5dB to the QPSK receivers, and additional data in equivalence of 3 channels for a penalty of 0.5dB. We note that these it rates are heuristic estimates. Similar estimates can also e derived y using the channel capacity of the AWGN channel [9, p.08] 8
Local Area local data local data New modulator Local Repeater Old Local Area local data local data New modulator Local Repeater Old Local Area 3 local data 3 local data 3 New modulator Local Repeater Old C = log ( + SNR). (8) However, the actual it rate of the secondary information depends on selections of the channel coding and decoding algorithms. The rate of the secondary channel may e increased y using larger values for λ, at the expense of more penalties for the QPSK receivers. A recommended tradeoff would e to use a small the value of λ when the upgraded system is initially introduced, in order to reduce the penalty to the QPSK receivers. As more new generation receivers are deployed, and the originally designed receivers are gradually phasing out, the value of λ can e increased to raise the it rate of the secondary information. V. LOCAL INFORMATION Figure 8 Transmission of local information In this section, we descrie how the hierarchical modulation can e used to transmit local information efficiently in a network with local repeaters. In satellite roadcast systems such as digital satellite radio, signals are transmitted to moile receivers. In some areas such as metropolitan regions, signals from the satellites are weak due to lockage such as tall uildings. It is a common practice to set up terrestrial repeaters in places where it is difficult to receive satellite signals. The terrestrial repeaters transmit the same data content as the satellites, ut with possily a different modulation format. With the aid of hierarchical modulation, the terrestrial repeaters can e used to provide local information that is of only interests to the local region covered y a repeater. Examples of local information include local news, traffic, weather and advertisements. Figure 8 shows how the hierarchical modulation can e used to provide the local information efficiently in systems with a gloal network and local repeaters. The local information can e modulated in the secondary hierarchy, either as a part, or the entirety of the secondary information. In this way, the local information can e generated locally, and it needs not e carried in the gloal network, hence improving the andwidth efficiency of the gloal network. When all local repeaters transmit in a same frequency and, they form a single frequency network (SFN). Since local information content may e different from repeater to repeater, there exist interferences 9
etween neighoring repeaters. To comat the interferences, the secondary information may e modulated with orthogonal pulses such as CDMA. A small numer of orthogonal codes can e predefined, so that each repeater in a cluster of neighoring repeaters, such as in a single metropolitan area, uses a different code, and no repeaters of same code will share an overlapping coverage region. The hierarchical receivers must e designed to e capale of demodulating all codes, and e ale to decide which repeater s local information is used as output. VI. CONCLUSION The analysis shows that the hierarchical modulation can e used effectively to upgrade a digital roadcast system in response to oth the demand for higher it rate that is made possile due to advances in technology and coding algorithm development, and the need to e ackward compatile to the already deployed old receivers. The added value of the paper is the analysis of the impact that hierarchical modulation has on the already deployed receivers versus the secondary information itrate. The hierarchical parameter λ can e used y the system operator to control the tradeoff etween the penalty to the already deployed receivers and the it rate of the added secondary information. This paper has een concerned with the modulation scheme. Another important issue to consider in the system upgrade is the channel coding for the secondary information, which, together with λ, will ultimately determine the actual it rate of the secondary information. The issue of channel coding has not een addressed in this paper. Further studies are warranted to find an optimal channel coding in conjunction with the hierarchical modulation. rate, IEEE Trans. Inform. Theory, vol. IT-0, pp. 84 87, 974. [4] P. W. Wolniansky, G. J. Foschini, G. D. Golden, R. A. Valenzuela, V-BLAST: An architecture for realizing very high data rates over the Rich-scattering wireless channel, invited paper, Proc. ISSSE-98, Pisa, Italy, Sept. 9, 998. [5] T. Cover, Broadcast channels, IEEE Trans. on Inform. Theory, Vol IT-8, pp. -4, 97. [6] K. Ramchandran, A. Ortega, K. M. Uz, and M. Vetterli, Multiresolution roadcast for digital HDTV using joint source/channel coding, IEEE J. Select. Areas Commun., Vol, 993. [7] P. Vitthaladevuni and M. Alouini, BER computations of 4/M-QAM hierarchical constellations, IEEE Trans. Broadcasting, Vol. 47, pp. 8-39, 00. [8] C. Berrou, A. Glavieux and P. Thitimajshima, "Near Shannon limit error-correcting coding and decoding: turo-codes", Proc. of IEEE ICC '93, Geneva, pp. 064-070, May 993. [9] E. Lee and D. Messerschmitt, Digital Communication, nd Ed., Kluwer Academic Pulishers, Boston, 994. [0] Alexander Schertz and Chris Weck, Hierarchical modulation - The transmission of two independent DVB- T multiplexes on a single frequency, EBU Technical Review, April 003. [] Gerhard Kramer, Private communication, 003. [] ETSI EN 300 744 V.4., European Standard (Telecommunications series) Digital Video Broadcasting (DVB); Framing structure, channel coding and modulation for digital terrestrial television, 00. REFERENCES [] M. Lamarca, S. Sala and A. Martinez, Iterative decoding algorithms for RS-convolutional concatenated codes, Proceedings of International Symposium on Turo Codes, Brest, France, Septemer -5, 003. [] A. J. Viteri, An intuitive justification and a simplified implementation of the MAP decoder for convolutional codes, IEEE J. Select. Areas Commun., Vol. 6, No., 998. [3] L. R. Bahl, J. Cocke, F. Jelinek, and J. Raviv, Optimal decoding of linear codes for minimizing symol error 0