EFFECTS OF PHYSICAL LAYER IMPAIRMENTS IN DIGITAL VIDEO BROADCASTING BY SATELLITE. Elias Nemer and Ahmed Said

Transcription

1 EFFECTS OF PHYSICAL LAYER IMPAIRMENTS IN DIGITAL VIDEO BROADCASTING BY SATELLITE Elias Nemer and Ahmed Said Advanced Technology Office. Consumer Electronics Group. Intel Corporation 350 E. Plumeria Drive, San Jose, CA 9534 U.S.A phone: (408) , fax: (408) ; ABSTRACT Digital Video Broadcasting via satellite is being widely deployed worldwide. The first generation European standard (DVB-S) is a QPSK based system that supports up to 40 Msymb/s transmission rates. This system is also a component in the North American (ATSC) standard and one of the 4 subsets of the ITU standards. Higher order modulations (from 8PSK to 3 APSK) are incorporated in the recently released second generation specification (DVB-S). Physical layer impairments on the transmitreceive link include those introduced by the channel, such as reflections and interference, as well as those induced by various components in the system, such as tuner I/Q imbalance or amplifier non-linearity. These impairments degrade the received SNR and may affect the convergence behavior of the various receiver loops. The paper provides a qualitative and quantitative description of the system and channel impairments and their effects on the received cluster SNR. Implications and what countermeasures should be considered in the receiver design are also discussed.. INTRODUCTION Digital video broadcasting delivers crystal clear pictures and CD-quality sound, creating a true hometheatre experience for the consumer. A number of standards for terrestrial [] [4] satellite [5] [8] and cable broadcasts [0] have been defined over the past few years. DVB by satellite is being widely deployed worldwide and is already evolving to a second generation (DVB-S [7] ) to accommodate higher bandwidth efficiency, more elaborate service types and a better quality through new concepts such as hierarchical modulation. The general interference and attenuation environment for a satellite link in shown in Figure. An earth station transmits on the uplink with a given power P. Other stations destined to the same transponder and interfering sources on the uplink may also be transmitting towards the same receiving antenna. At the transponder, the aggregate of all transmissions is bandpass filtered and pre-amplified. Figure : Total channel path in DVB by satellite The frequency converter in the transponder shifts the uplink transmission to the appropriate band destined for the downlink. The TWTA amplifies the signal with a gain β and the signal is radiated with a total power: Psat = Pg G. On the downlink, weather effects, ST such as rain and scintillations, may attenuate the signal. Moreover, interference from adjacent satellites and ground stations can also degrade the broadcast signal. At the receiver, the LNA shifts the incoming signal from GHz band (Ku band) to about to GHz. The tuner demodulates the signal and provides the (analog) I and Q components to the rest of the (digital) receiver. This last stage further provides the digital down sampling, symbol timing adjustment, equalization, and any further required processing (error correction and de-multiplexing) to generate the final MPEG packets. The Carrier to noise ratio in db-hz at the IF output of an earth terminal is function of the transmitted power, back off levels, total signal path loss and system temperature [9] : C N 0 G = P max B 0 + βp + κ ζ L T Equ

2 P is the maximum peak effective radiated power in max dbm. B is the back off amount in db of the total average 0 satellite power i.e. β P is the percentage in db of the available average power to be used for the carrier of interest (e.g. 0% corresponds to 0 db). G / ξ is the ratio of effective receiving antenna gain to received system noise temperature. κ is the boltzman constant, -98 dbm/k-hz; and L is the total loss, including free space, T antenna pointing loss, rainfall, etc to that terminal. The above C/N 0 ratio thus accounts only for additive noise, as well as the radiated signal strength and the attenuation throughout the channel and the system. In typical manufacturers and service providers specs, separate ratios are provided for such factors as the added interferences from adjacent satellites, the intermodulation products, the cross-polarization noise and others, as discussed in the sections below.. CHANNEL IMPAIRMENTS.. Free-space Loss The loss in signal strength through free space is function of the frequency and the distance between the satellite and the receiver: L f = log0( f ) + 0log0( r). For geo-synchronous satellites, orbiting at 35,784 km and a frequency of GHz, the loss is: L f = 05. 5dB. as the geographical location, the angles and latitude of the antennas, the outage probability desired, etc..3. Non-linearity of TWT Amplifiers is due to the analog filters and amplifiers on the path from the earth station to the satellite and down to the (user) receiver. The dominant ones are due to the traveling tube amplifier (TWTA) and may be represented in terms of amplitude and phase effects, termed AM-AM and AM-PM between the input and the output. A general PSK signal at the input of a non-linear system x( t) = r( t) cos[ wt +θ] yields an output of the form y( t) = A[ r( t) ] cos{ wt +θ + φ [ r( t) ]}, where the two non-linear functions are expressed in terms of the input envelop r(t) through a set of curves, or in a parameterized form such as the commonly used Saleh model [4] : a a r rr φ A( r) = ; φ( r) = with a typical set + b r + b r r ar =.587; br =.57; a φ = 4.033; bφ = The amount of non-linearity depends on the operating point, or input back off (IBO) relative to the point of saturation. Usually, or more sets of curves are given, depending on whether or more carriers are being fed into the same TWTA amplifier. The effect on a QAM constellation is significant for higher order modulation, such as 6 QAM, but less so for the QPSK case (figure ) φ.. Rain and Scintillation Rain cells have a dual effect on the C/N ratio:. They produce an attenuation at the carrier level, thus contribute to the total loss L T (in Equ ).. They increase the antenna (and the system) temperature by as much as 00K relative to a clear day, resulting in degradation in the G / ξ ratio. In 0 30 GHz, the attenuation coefficient variation is nearly linear. Analytically, the amplitude attenuation is log normal distributed (i.e. slow and flat fading), with the density function given by [] : ( y m ) y. py ( y) = exp σ σ y π y Scintillations are distortions by small-scale irregularities of the refractive index in the path. The amplitude (attenuation) of scintillation is log normal distributed and over short durations (up to minute), the variance is considered constant. As with rain, the density function is Gaussian. While this analytical model is mostly useful for simulation, for the sake of link budget computations, the actual attenuation is determined by a recipe-like guidelines, such as the one provided in IESS-308, Appendix J, [] and is function of several factors, such Figure : Effect of non-linearity on QPSK and 6 QAM.4. Co-Channel Interference Co-channel interference is an interfering QAM signal that may overlap partially or completely in the same BW as the desired signal. This type of interference may be caused by a number of factors: Side lobe interference from a ground station. Cross-polarized signal from the same satellite. Inter-modulation products caused by the nonlinearity of the TWTA amplifiers when or more carriers are used in the same amplifier. Signal from ground microwave link in the same band

3 Adjacent satellites whose antenna side lobes may radiate across the spectrum of interest. Table below from the Optus satellite specifications [5] illustrates typical co-channel interference values. Table : Co-channel channel interference Parameters Clear sky Faded Uplink Faded downlink Intermod noise C/Im 7 db 3 db 7 db Downlink x-pol C/I Downlink adj satellite C/N Inter-modulations products are generated when more than carrier is used per amplifier. The amplitude of these products is negligible at low power and proportionally higher near saturation. As a rule of thumb for the case of carriers per amplifier, the level of intermod products (in dbc) is given by: IM dbc = ( * TOPB + ) db, where TOPB is the total output power back off (from saturation)..5. Tuner I/Q imbalance An analog tuner is commonly used in the front end to down-convert the signal (in frequency) and to provide a first stage of demodulation that yields the baseband quadrature (I and Q) components. Due to imperfect analog multiplications and filtering, the quadratures are off-balance, in both phase and amplitude. The corrupted quadratures (I and Q ) may be written in terms of the un-corrupted (hypothetical) quadratures (I and Q) as well as the magnitude imbalance (ε ) and phase imbalance ( ϑ ) [3] : I ' + ε = Q' 0 ϑ cos 0 ε ϑ sin ϑ sin I ϑ Q cos For MPSK modulations, the I/Q imbalance causes a warping of the (otherwise round) constellation (figure 3). The effect is more pronounced for square constellations such as the 6-QAM, and is less significant for the QPSK case. For the multi-ring PSK constellations specified in [7], the effect is also pronounced since the distance between points on different rings is significantly altered. Figure 3 : Effect of I/Q imbalance on 8PSK and 6QAM.6. Mutipaths Since there is often a line of sight reception in DVB-S, the reflections due to signal bouncing off the neighboring obstacles (buildings, etc...) have negligible power compared to the main signal. However, in contexts where a (cable) distribution system is used, micro-reflections due to the cable and the connections are significant enough to cause noticeable signal degradation A Satellite Master Antenna System (SMAS) is defined in [6] as one which is intended for the distribution of television and sound signals to households located in one or more adjacent buildings. It is a means for sharing the same resources among several users for satellite and terrestrial reception combined micro-reflections model (shown in the figure below). In the model specified, micro-reflections between devices in consecutive floors, between head-end and the first device, and between tapoff and user outlet, are specified. In the aggregate model, the shortest echo is about 30 nsec and the longest about 40 nsec. The effect on higher order constellation is significant as may be seen in figure 4 below, where echoes at 38, 55, 85 and 5 nsec, with attenuation according to the echo path profile in [6], are simulated for an 8PSK constellation. The graph on the right shows the case when AWGN only at 5 db is added to the signal. Figure 4: Effect of reflections vs. AWGN only for 8PSK

4 .7. Orbit Inclination & Doppler Shift The satellites used in broadcast services are geostationary and are almost equatorial: During the first 5 or so years, the satellite is maintained at equatorial orbit +/ deg. Afterwards, north-south correction ceases and the orbit inclination builds in a linear fashion at 0.8 deg / year up to 3 deg. The impact of an inclined orbit is two-folds: The linear velocity (towards and away from the receiving earth station) causes a Doppler shift in carrier frequency, which affects the overall frequency tracking accuracy. The total path length is (slowly) varying with time and causes a change in the total path delay. The linear velocity of the satellite may be determined analytically, given the inclination angle, the height of the satellite and the latitude of the user station. For example, it may be shown that for the case of an end user at the equator (figure 5), the linear velocity is given byv = ( D H ) / 6 km / h, where H is the height of the satellite and D is the maximum distance to the satellite from the observer and is given by D = R{ cos( I )} ( R + H ) + H, with R the radius of the earth, and I is the inclination angle of the satellite. For a height of 35,784 km, a radius R of 6378 km and a 5 deg inclination, V = 4.76_ km/ h. In practice, satellite manufacturers and service providers supply a table of maximum frequency shift (for instance the figures for the VA and VI satellites in [3] ). I In the shown position, the distance to the satellite is D, When the earth (and satellite ) are ¼ turn, the distance becomes H I R Thus motion speed to user: H ( D H ) V = km / h D 6 Equator H : height of satellite = 35,784 km R : radius of earth = 6378 km D : distance to satellite from an observer observer on equator I : inclination angle of satellite orbit R R H N D N M Equator M = ( R + H )sin( I) N = ( R + H )cos( I) N = R + N N = N R { cos( I) } ( R + H ) D = R + H Figure 5 : orbit inclination effects.8. Phase Noise Phase noise arises from various sources: The frequency conversion process in the satellite including the TWTA phase noise. The frequency converters and amplifier (LNA) at the user site that bring the signal from to GHz. The tuner itself where the down-conversion to baseband is done. The specifications of a typical tuner phase noise are given in the table below. Table : Tuner phase noise typical case Offset freq KHz Phase noise dbc/hz.9. Frequency Error Frequency errors in the received baseband I and Q quadratures may be caused by a variety of sources along the path from the earth station to the user receiver. Two sources are worth noting: The frequency conversion process in the satellite (uplink to downlink freq). This is usually limited to +/- 5 khz. The frequency converters/amplifier at the user site (LNA) that brings the signal from to + GHz. The oscillators in the tuner 3. SIMULATION RESULTS A DVB-S transmitter/receiver system using QPSK, 8PSK 6APSK, 3APSK and 6-QAM modulation according to the specifications in [] [5] [8] is simulated using Matlab. The symbol rate is set to 30 Msymb/s and a sampling rate of 3-4 samples per symbol is used. A baseband channel model is assumed and various combinations of the following impairments are added: I/Q imbalance, withε = 0.5; ϑ = 5 deg. Up to 4 echoes at 36, 55, 85 and 55 nsec according to the echo path profile defined in [6]. Additive white Gaussian noise at various E b E. 0 Non-linearity with a back off point of 0 db and 5dB. Frequency offset with f [ 9 : 5]kHz. An ideal receiver consisting of a matched Nyquist filter and a down sampler is used. The cluster SNR of the received soft symbols is calculated for each impairment case. The results for a QPSK constellation are shown in Table 3 below. Table 3 : impairments effects on SNR for QPSK Impairments Parameters Cluster SNR AWGN Eb/No = 7.5 db 0.5 db

5 Echoes at {36, 55, 85, 55} nsec AWGN; AWGN; Reflections; ; IQ Imbalance AWGN; Reflections; ; I/Q Imbalance Non-linearity; Freq Offset ; ; I/Q Imbalance; Freq Offset 6.5 Back-off = 0dB 6.94 From the above 3 cases Amplitude = 0.5 phase = 5 deg From above cases 6.5 Freq offset at Fs/400 From all above cases 5.38 Table 4 compares the cluster SNR for a QPSK and an 8PSK constellation for given impairment scenarios. In each case, the same combination of impairments is used for both, except that the level of AWGN is different for the. The following impairment set is used: I/Q imbalance, withε = 0.5; ϑ = 5 deg. echoes at 55 and 85 nsec as defined in [6]. Additive white Gaussian noise at E b E0 = 7. 5 db for QPSK and E E = b db for 8-PSK. Non-linearity with a back off point of 5dB. Frequency offset with f = 9 khz. Impairments 5. Table 4 : Cluster SNR for QPSK and 8-PSK Eb/ No QPSK Cluster SNR Eb/ No 8PSK Cluster SNR AWGN db db AWGN, IQ Imbalance I/Q Imbalance Non-linearity Freq Offset I/Q Imbalance Freq Offset IMPLICATION FOR RECEIVER DESIGN Equalization: an equalizer is required to counter the effect of the echoes in the antenna cable distribution. Given the echo profile, it is estimated that a T-space equalizer of about 4 taps will be sufficient to compensate for all echoes. I-Q imbalance: simulation show that the convergence of the equalizer is negatively impacted by the I/Q imbalance and yields a worsened constellation at the output of the receiver. Thus, an explicit compensation scheme possibly at the front end of the receiver is needed to remove the effect of the I/Q imbalance. Frequency and phase offset correction: Frequency offsets are large enough to cause errors, particularly for 8PSK and 6QAM constellation. Therefore a correction scheme is required. Given the large offset amount, it is common to use compensation loops: the first providing a coarse (and blind) correction and a second loop for a fine phase correction. Non-linearity: for the case of QPSK, the effect of non-linearity is negligible and no explicit remedy is needed. For the case of a 6QAM constellation, either non-linear equalization is needed or a slicer map that will account to the warping effects is needed. A typical DVB-S receiver block diagram is shown in figure 6 below. Figure 6 : DVB-S receiver block diagram

6 5. CONCLUSION In this paper, we provided a qualitative and quantitative description of the physical layer impairments in a DVB-S system and their effects on the received cluster SNR. The analysis was illustrated with comparative results for QPSK and 8PSK constellations for a number of impairment combinations. The results show that both constellations are susceptible to most of the impairments and that specific countermeasures are needed particularly for channel multi-paths, I/Q imbalance, and frequency offset. The effect of multi-paths and I/Q imbalance is more pronounced for 8PSK and 6QAM, though degradation is noticeable even for QPSK. The effect of non-linearity is not significant for the case of QPSK, but is pronounced enough for the case of 8PSK and 6 QAM, requiring either proper (nonlinear) equalization or a change in the decision boundaries.. REFERENCES [] A/53. ATSC standard: digital television standard, revision B. Advanced Television Systems Committee (ATSC). Aug 00. [] A/80. Modulation and coding requirements for digital TV applications over satellite. ATSC standard July 999. [3] M. Buchholz, A. Schuchert, R. Hasholzner, Effect of tuner I-Q imbalance on multicarrier-modulation systems, Proc of the 000 conf on Devices, Circuits, and Systems, March 000. Vol., pp. T65/ T65/6. [4] EN Digital video broadcasting. Framing structure, channel modulation for digital terrestrial television. ETSI V [5] EN Digital video broadcasting (DVB). Framing structure, channel coding and modulation for / GHz satellite services. ETSI. V [6] EN Digital video broadcasting (DVB) satellite master antenna television (SMATV) distribution systems. ETSI. V [7] EN Digital video broadcasting: second generation framing structure, coding and modulation systems for broadcasting, interactive services, news gathering and other broadband satellite applications. ETSI. DVBS-74r [8] ITU-R BO.56. Common functional requirements for the reception of digital multi-programme television emissions by satellite operating in the / GHz frequency range. 00 [9] J. Spilker. Digital Communications by Satellite. Prentice Hall [0] ITU-T J.83. Digital multi-programme systems for television, sound and data services for cable distribution [] J.P. Choi & V.W. Chan Predicting and adapting satellite channels with weather-induced impairments. IEEE trans. on aerospace and electronic systems, Vol. 38, issue 3, July 00, pp [] Intelsat Earth Station Standards (IESS) 308. Rev 0, Feb 000. Performance characteristics for intermediate data rate digital carriers using convolution encoding / viterbi encoding and qpsk modulation. [3] Intelsat Earth Station Standards (IESS) 4. Rev 4, Oct 00. Requirements for earth stations accessing Intelsat VA & VI Satellites having higher than nominal orbital inclination. [4] A.A.M. Saleh, Frequency-independent and frequency-dependent non-linear model of TWT amplifiers, IEEE trans. Commun. vol. 9, pp , Nov 997. [5] Optus Satellite Network Designer s Guide.