DVB-S2 Satellite Experiment Results

DVB-S2 Satellite Experiment Results Nicolas Girault 1, Olivier Smeyers and Riccardo de Gaudenzi European Space Agency, Noordwijk, The Netherlands Cyrille Moreau 2, Eric Alberty and Cédric Le Guern EADS Astrium Satellites, Toulouse, France and Andre Bigras 3, Geoff Dane, Surinder Singh and Peter Megyeri Telesat, Ottawa, Canada With a large number of profiles and options, the DVB-S2 standard offers several possibilities to move towards the Shannon capacity bound and more in general to reduce the delivery cost per bit in real system scenarios, compared to previous generations of air interface. The information bit delivery cost with high quality of service is a key factor of success for both broadcast and interactive high throughput satellite systems. This paper presents the main results and findings of extensive European Space Agency (ESA) funded DVB-S2 test campaigns performed by satellite on three complementary test platforms, addressing broadcast, professional and interactive profiles. Different satellite transponders operating in C-band, Ku-band and Ka-band and located in Europe and in Canada have been used encompassing both single beam and multi beam satellite coverage. The performance of high order modulations has been measured, as well as the effects of payload impairments with and without mitigation techniques such as modulator pre-distortion and receiver equalisation, the impact of phase noise, etc. Particular effort has been dedicated to investigate the performance of Adaptive Coding and Modulation (ACM) in real and emulated conditions. The paper also provides recommendations concerning different tradeoffs to be made when operating DVB-S2 carriers, like optimization of ACM margins, amplifier back-off setting and insertion of pilot symbols. C/N C/(N+I) E S /N 0 IBO OBO R S Nomenclature = roll-off factor = carrier-to-noise power ratio (N measured in a bandwidth equal to symbol rate) = carrier-to-(noise+interference) ratio = ratio between the energy per transmitted symbol and single sided noise power spectral density = input back-off = output back-off = symbol rate corresponding to the bilateral Nyquist bandwidth of the modulated signal I. Introduction NTRODUCED in 2005, the DVB-S2 standard 1 has been rapidly adopted in satellite telecommunications, both for I consumer and professional services. The standard is nowadays commonly used for direct-to-home delivery of bandwidth hungry broadcasting services such as High Definition (HD) television and emerging 3DTV. DVB-S2 is also widely used in Ka-band interactive satellite networks (Telesat Anik F2/F3, Eutelsat KaSat, Avanti Hylas, ViaSat WildBlue, etc) for broadband services as well as for contribution links. Following the development of an early DVB-S2 test-bed supporting ACM 2,3 in 2004, where extensive tests have been performed in laboratory 1 Ground Segment Engineer, Telecommunication Technologies, nicolas.girault@esa.int 2 Telecom System Engineer, Astrium Telecom Systems Department, cyrille.moreau @astrium.eads.net 3 Manager, Advanced Systems Engineering, Space and Network Engineering, abigras@telesat.com 1

environment, in 2006 ESA launched the DVB-S2 Satellite Experiment project 4. The main objective of the activity was to perform an in-depth verification and optimisation in real satellite conditions of the many new features present in the standard, such as enhanced physical layer coding, high order modulations, associated pre-distortion techniques and ACM, through extensive test campaigns consisting of field trials and measurements. The main phases of the project were the following: Development of test equipment and facilities covering the different profiles of the DVB-S2 standard (broadcast, professional and interactive) to be connected to existing satellite space and ground segments. Functional verification of the DVB-S2 features by satellite for a number of configurations and applications of practical interest to satellite operators. Satellite test campaigns with C-band, Ku-band and Ka-band transponders, focussed on the performance aspects, to assess in particular the efficiency and availability of the advanced modulation schemes, LDPC/BCH coding and ACM adaptation. Measurement of the performance of the DVB-S2 broadcast equipment combined with SDTV and HDTV programmes and advanced video compression techniques like MPEG-4 / AVC. II. Description of the Test-Beds Three platforms corresponding to the main profiles of the DVB-S2 standard have been put in place for the experiments: broadcast, professional and interactive. Each platform could either be connected to the satellite channel, to perform realistic tests in a given environment, or to a channel simulator in a lab controlled environment (L-band loop-back). The two channel paths could also be combined, for instance to generate artificial fading series on top of a real satellite link. The main characteristics and differences of the platforms are shown in Table 1, together with the reference link budget. Broadcast Profile Professional Profile Interactive Profile Space Segment Telesat Eutelsat Telesat Satellite Anik F1R Atlantic Bird 3 Nimiq 4 Orbital location 107.3 West 5 West 82 West Geographical coverage Canada Europe CONUS Frequency band C-band Ku-band Ka-band Downlink frequency 3800 MHz 11013 MHz 18700 MHz Transponder bandwidth 36 MHz 36 MHz 54 MHz Ground Segment Product readiness Commercial ESA Prototype Pre-commercial Teleport Ottawa, Canada Rambouillet, France Ottawa, Canada Applications HDTV broadcasting Professional app. Consumer app. DVB-S2 Parameters (Default) Symbol rate 25 Mbaud 10 Mbaud 9.2 Mbaud Roll-off factor 20% 25% 20% Mode CCM - VCM CCM - ACM ACM Nb MODCOD tested / used in ACM 28 28 / 22 28 / 21 Frame length (bits) 64800 (normal) 64800 (normal) 16200 (short) Pilot symbols On On On Forward Link Budget (Clear Sky) Transponder OBO 0.35 db 2.3 db 3 db Receive antenna diameter 2 m 1 m 18 m 3.5 m Receiver C/(N+I) 7.9 db 16.2 db 21.7 db 24.9 db Table 1. Platforms summary The main DVB-S2 parameters in Table 1 represent the default values subsequently used for the tests (unless specified otherwise) and the reference link budget for clear sky conditions. Since dedicated transponders were allocated to the experiments, it was possible to configure the spectrum with one or several carriers per transponder. The Input and Output Back-Off (IBO/OBO) operating points of the transponder could be changed by the operator for specific testing. Receive antenna sizes were oversized to obtain high link budget margin and therefore allow transmission of highest spectral efficient DVB-S2 MODCOD, or to allow adding channel effects a posteriori for sensitivity tests on a wide E S /N 0 range. Realistic receivers Outdoor Units (ODU) have also been used. It was possible to increase E S /N 0 by reducing symbol rate R S at constant satellite EIRP (equivalent isotropically radiated power). Eventually, all 28 MODCOD of the standard have successfully passed by satellite on each platform. 2

A. Broadcast Profile The broadcast profile test platform was based in Ottawa (Canada) and included a range of test capabilities. The architecture of the test bed allowed testing of the DVB-S2 commercial modulator and demodulator in an IF or a satellite loop-back. The setup was capable of providing varying C/N values, impose time series and insert different phase noise masks on the modulated carrier. The test-set up also incorporates test points to facilitate monitoring of the signal spectrum in the transmit and receive chains. B. Professional Profile Figure 1. Broadcast profile test-bed Hub to Client PC Gateway Application PC PC Traffic Traffic Generator & Analyser KVM SNMP GPS Receiver KVM Gateway NCC/NMS MF/TDMA Receiver Scheduler SNMP GPS Receiver DVB-S2 DVB-S2 Modulator Consoles Amplifier Up-Conv. Interface Module Module sat lab Channel A Channel A Simulator KVM SCPI DVB-S2 Carriers SNMP sat lab A LNB LNB A Noise Noise Source GPIB Spectrum Analyser GPIB Console DVB-S2 /RCS /RCS Terminal Noise XML Noise Source GPIB DVB-S2 /RCS /RCS Terminal SNMP XML Console Spectrum FTP Analyser GPIB Toulouse Site Rambouillet Site From Gateway PC RS-232 MOXA MOXA RS-232 MOXA MOXA Application VNC / Remote Remote PC PC Desktop Traffic Traffic SNMP Generator & Analyser Terminal Application PC PC Traffic Traffic Generator & Analyser KVM SNMP NTP Internet VPN Tunnel Switch Router Firewall Firewall Router GPS Receiver Switch KVM KVM Switch Switch Control & M&C Monitoring System NTP System Figure 2. Professional profile test-bed 3

The professional profile test platform was encompassing two sites (Rambouillet and Toulouse both in France) connected to Eutelsat Atlantic Bird 3 satellite. The platform consisted of the IDU and ODU transmission equipment under test and the test equipment. The main site was the Eutelsat teleport in Rambouillet, where most of the equipment from the former ESA s ACM Modem project have been installed. Additional equipment have been manufactured and procured for the platform monitoring and control, which can be controlled from any site using the extended LAN via the VPN. The test-bed detailed design, including demodulator algorithms for channel estimation, is described in the ACM Modem article 2. C. Interactive Profile The interactive profile test platform used a pre-commercial DVB-S2/RCS system (hub and remote terminals). The test-bed architecture, similar to the two other test-beds, was constructed to permit seamless insertion of test devices (computers, test sets etc.) both on the hub side and at the output of the IDU s as required. With such a wide and varied range of test requirements it was essential that test equipment be easily, logically and/or physically inserted into the signal path. Testing was implemented both at IF loop-back and at RF loop-back using the Telesat Nimiq-4 satellite (Ka-band). Figure 3. Test racks at Telesat headquarters Figure 4. Nimiq-4 beam coverage over Canada III. Main Results and Recommendations A. Link Performance in Static Conditions The end-to-end link performance has been characterized in static conditions for each MODCOD on each platform. White Gaussian noise was added in front of the demodulator, to generate controlled E S /N 0 levels. E S /N 0 measurements presented in this paper were either performed within the demodulators, based on internal estimators, or using an external spectrum analyser (channel power method). Figure 5 shows DVB-S2 baseband frame error rate (FER) statistics as a function of E S /N 0 measured on the professional profile in laboratory environment (L-band loop-back). It is observed that the FER vs. E S /N 0 curves are very steep below 10-2 (typically 1 decade per 0.1 db), especially for the low MODCOD. It was noted however in subsequent tests by satellite that the slope was not so steep for high MODCOD in the presence of link impairments. It can be observed from Figure 5 that E S /N 0 spacing between MODCOD performance curves is not uniform. There are overlap areas between modulations, where E S /N 0 thresholds required for QEF behaviour are separated by only 0.1 or 0.2 db. In other regions, gaps up to 1.4 db are measured, like between 32APSK-5/6 and 32ASPK-8/9. Using a finer MODCOD set would reduce the quantisation step with immediate gain. Assuming that C/(N+I) link budget distribution is uniform in the interval between two consecutive E S /N 0 thresholds, the average gain of introducing an intermediate MODCOD can be estimated as half of the distance between the thresholds. For instance, creating a QPSK-5/9 MODCOD in between QPSK-1/2 and QPSK-3/5 would reduce the quantisation loss from 0.6 db to 0.3 db approximately, which would equivalently improve spectral efficiency by 3.8% in this region. 4

Frame Error Rate after Decoder 1E+0 1E-1 1E-2 1E-3 1E-4 1E-5-3 -2-1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Demodulator Es/No in db QPSK-1/4 QPSK-1/3 QPSK-2/5 QPSK-1/2 QPSK-3/5 QPSK-2/3 QPSK-3/4 QPSK-4/5 QPSK-5/6 QPSK-8/9 QPSK-9/10 8PSK-3/5 8PSK-2/3 8PSK-3/4 8PSK-5/6 8PSK-8/9 8PSK-9/10 16APSK-2/3 16APSK-3/4 16APSK-4/5 16APSK-5/6 16APSK-8/9 16APSK-9/10 32APSK-3/4 32APSK-4/5 32APSK-5/6 32APSK-8/9 32APSK-9/10 Figure 5. FER vs. E S /N 0 for each MODCOD on the professional profile (laboratory configuration) When looking at the performance for Quasi Error Free (Figure 6), E S /N 0 degradations measured are ranging from 0.1 to 0.5 db compared to the theoretical performance from the DVB-S2 standard (chapter 6). The unexpected degradation of 0.5 db for MODCOD 12 (8PSK with FEC rate 3/5) has also been observed on the other profiles with disparate terminal equipment, presumably due to the inherent LDPC rate 3/5 code weakness in the presence of channel impairments (e.g. phase noise). When transmitting by satellite, in the linear region of the TWTA, additional degradations caused by channel impairments have been measured most likely due to the linear distortions occurring in the satellite transponder and RF front-end. Besides, demodulators located behind the small and large antennas have shown similar performance in terms of E S /N 0 degradation. It must be noted that signal levels are more variable when transmitting over satellite channel, which cause E S /N 0 measurement uncertainties of typically 0.1 to 0.2 db. Es/No Degradation in db 1.3 1.2 1.1 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 Measurements over Satellite Measurements in Laboratory QPSK 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 MODCOD Index Figure 6. E S /N 0 Degradation in laboratory and by satellite at Quasi Error Free. QEF target corresponds here to a frame error rate (FER) of 5*10-4. E S /N 0 reference performance is the DVB-S2 standard. 5 8PSK 16APSK 32APSK

B. Performance of ACM by Satellite The ACM mechanisms of the DVB-S2 standard allow to dynamically adapting the transmission spectral efficiency depending on C/(N+I) value experienced by the user terminal while keeping the occupied bandwidth constant. ACM requires a return link channel normally by satellite for sending user terminal C/(N+I) measurement feedback. It is particularly efficient for links with potentially high fading such as Ka-band or tropical regions. The potential performance advantage of ACM systems have been assessed through multi-dimensional link budget simulations in the former ACM Modem ESA contract 2.Compared to static systems using a fixed fading margin, typically dimensioned for a link availability of 99.5% or better, ACM allows each individual link to adapt its transmission mode to the current C/(N+I) conditions (spatial and temporally variant) thus maximizing the system spectral efficiency. The other improvement brought by ACM concerns the link availability which can be guaranteed without recurring to worst-case static link margins as it is the case in not ACM-based systems. The correct functioning of ACM was confirmed by satellite on the professional and interactive profiles, respectively in Ku-band wide-beam and Ka-band spot-beam configurations. An example of adaptation to a strong natural fading event is given in Figure 7. The conditions have been described by the local weather office as storm with thunder, heavy rain and water accumulating on ground. The wind was moving in perpendicular direction of the satellite link at about 20 km/h. The recorded C/(N+I) time series and associated MODCODs for the two sites are shown in Figure 7. It is important to note that, since the gateway and terminals in Rambouillet are co-localised, the fading event impacts both to the uplink and on the downlink. A correlation is clearly observed between Rambouillet and Toulouse small antennas, by comparing the C/(N+I) curves, meaning that the C/(N+I) at the two different user terminal locations is impacted by the feeder link fading event. More precisely as the satellite transponder ALC was deactivated (fixed gain mode), uplink power level variations impacted the downlink through the amplifier AM/AM response. Toulouse terminal was only impacted by the fading event on the uplink, the downlink being under better and more stable link conditions. 19 C/(N+I) received and Es/No thresholds in db 18 17 16 15 14 13 12 11 10 C/(N+I) in Rambouillet on Large Antenna Receiver C/(N+I) in Toulouse on Small Antenna Receiver C/(N+I) in Rambouillet on Small Antenna Receiver (ACM Loop) Es/No Threshold for the Active MODCOD (ACM Loop) 5 10 15 20 25 30 35 40 45 Time in minutes Figure 7. Example of C/(N+I) impact of a storm event as perceived by 3 terminals operating in Ku-band, and consequent ACM MODCOD selection. E S /N 0 threshold levels are shown for the small antenna receiver in Rambouillet, and must always be below C/(N+I) in order to avoid traffic losses. When considering the satellite environment and the multiple-terminal aspect, it must be emphasised that: The accuracy of the demodulator E S /N 0 estimator is of prime importance to optimize the ACM decision. Inaccuracies and variations in this measurement require a specific ACM margin, which causes a loss of spectral efficiency. The ACM margins have to accommodate scintillation effects, which fluctuations are usually faster than the ACM loop response time, as well as possible variations of the signal level through the satellite channel. Such effects should not be neglected otherwise ACM margins could be insufficient and outage events may occur. 6

C/(N+I) variations depend on C/(N+I) actual values, as measured on several Ku and Ka satellite links 5. ACM margins can therefore be optimized per MODCOD, taking advantage of the lower variability observed on high C/(N+I) values. C/(N+I) values and its variations are location dependent, due in particular to different satellite EIRP, different satellite antenna C/I over the coverage in the case of multi-beam satellite, terminal dish pointing errors, and possibly different types of terminals being used. One approach for the ACM margins is to consider a single worst case value on the coverage. Another approach is to customise ACM margins per terminal, in a pre-determined way (geographical pattern) or automated way (self calibrated terminals). The optimisation of the TWTA operating point in the case of ACM is more complex than in the case of CCM, because non-linear degradations are dependent on both modulation and coding rate. In order to maximise the system efficiency globally, it is necessary to take into account the expected C/(N+I) distribution and corresponding MODCOD, and to optimise the TWTA operating point for the most commonly used MODCODs. This way, the relative loss of efficiency for the other MODCODs remains marginal. C. Phase Noise Impact The impact of phase noise, usually coming from the frequency conversion in the receiver LNB, has been measured on the professional profile in L-band loop-back configuration. Different phase noise masks have been synthesised (Figure 8) from the DVB-S2 Standard (Annex H.8) and from a range of commercial LNB products. The channel simulator has been used to inject phase noise on the link. Frequency Offset 100 Hz 1 khz 10 khz 100 khz 1 MHz 10 MHz -20 Phase Noice in dbc/hz -30-40 -50-60 -70-80 -90-100 DVB-S2 Typical Mask DVB-S2 Critical Mask DVB-S2 Typical Mask + 5 db TV LNB Professional LNB 1 Professional LNB 2 Professional LNB 3 Professional LNB 4 High Quality LNB -110-120 -130 Figure 8. Synthesised phase noise masks Figure 9 presents E S /N 0 degradations measured for the 28 MODCOD at a frame error rate of 10-3. The following results are highlighted: The impact of phase noise of the link performance primarily depends on the robustness of the MODCOD. High order modulations are more sensitive to phase noise because of their constellation more pronounced sensitivity to recovered carrier phase errors (lower distance between ring constellation points). Similarly more degradation is measured when LDPC code rate increases (less redundant FEC bits available). For the TV LNB, designed for DTH low-cost mass market, degradation varies between 0 and 1.2 db for QPSK and 8PSK modulations (normative modulations for broadcasting services). For higher modulations, normally not used in DTH, the degradation ranges from 0.5 to 1.6 db for 16APSK and goes up to 8.2 db for 32APSK rate 9/10. Performance degradation with professional LNB and high quality LNB, representative of interactive, DSNG and professional services, is generally very limited, even for high order modulations. E S /N 0 degradations averaged over the four professional LNB accounted for 0.04 db in QPSK, 0.1 in 8PSK and 16APSK, and 0.2 db in 32APSK. 7

10.0 Es/No Degradation in db 9.0 8.0 7.0 6.0 5.0 4.0 3.0 DVB-S2 Typical Mask DVB-S2 Critical Mask DVB-S2 Typical Mask + 5 db TV LNB Professional LNB 1 Professional LNB 2 Professional LNB 3 Professional LNB 4 High Quality LNB 2.0 1.0 0.0 QPSK 1/4 QPSK 1/3 QPSK 2/5 QPSK 1/2 QPSK 3/5 QPSK 2/3 QPSK 3/4 QPSK 4/5 QPSK 5/6 QPSK 8/9 QPSK 9/10 8-PSK 3/5 8-PSK 2/3 8-PSK 3/4 8-PSK 5/6 8-PSK 8/9 8-PSK 9/10 16-APSK 2/3 16-APSK 3/4 16-APSK 4/5 16-APSK 5/6 16-APSK 8/9 16-APSK 9/10 32-APSK 3/4 32-APSK 4/5 32-APSK 5/6 32-APSK 8/9 32-APSK 9/10 Figure 9. Performance degradation due to phase noise It can be concluded that measured performances with different types of LNB resulted to be better than with the typical phase noise masks from the DVB-S2 Standard Annex H.8, which can be considered as pessimistic. D. Satellite Non-Linearity Impact Figure 10 shows the measured performance for the broadcast profile in the case of a single carrier per transponder without and with dynamic pre-distortion on ground. For each MODCOD tested, the optimum TWTA operating point is found to minimise C/N total degradation, defined as the sum of non-linear degradation at QEF and OBO. Symbol rate was set to 20 Mbaud for all MODCOD, except for 32APSK modes where 10 Mbaud has been used for link budget closure reasons. The carrier roll-off was set to 0.25. 4 Es/No Total Degradation in db 3.5 3 2.5 2 1.5 1 Total Degradation without Non-Linear Pre-Distortion Total Degradation with Non-Linear Pre-Distortion Advantage of Non-Linear Pre-Distortion 0.5 0 QPSK-3/4 QPSK-9/10 8PSK-3/5 8PSK-3/4 Figure 10. Performance degradation due to the satellite non-linearity with and without pre-distortion 8 8PSK-8/9 16APSK-2/3 16APSK-3/4 16APSK-9/10 32APSK-3/4

As expected, the measurements show that the TWTA can be operated close to saturation for the quasi-constant envelope QPSK and 8PSK modulations (with constellation points located on a single circle). The resulting degradation is in the order of 0.5 to 1.0 db. Activating non-linear pre-distortion in the modulator gives a very small improvement for these low order modulations. Instead for 16APSK (dual ring constellation) and 32APSK (triple ring constellation) modulations, up to 3.4 db nonlinearity degradations are measured, for which pre-distortion is able to compensate up to 1.1 db. The results were based on a first generation commercial modulator and are likely to be improved further by more recent modulators with dynamic pre-distortion. In conclusion, when operating one carrier per transponder, it is recommended to use non-linearity precompensation on the feeder link to improve the end-to-end link budget, and therefore spectral efficiency. As mentioned before, the choice of the IBO/OBO operating points in ACM operation requires prior system analysis taking into account the non-linear degradations of each MODCOD. E. Receiver Equalisation Linear equalisation is usually implemented in satellite receivers to diminish the effects of linear distortions coming from the satellite channel (typically IMUX/OMUX filters group delay) and the receiver radio frequency front-end possible mismatches. Figure 11 shows how demodulator acquisition threshold can be significantly improved by using equalisation techniques. This test was performed over satellite at 20 Mbaud. Receiver equalisation can also be useful to limit a posteriori the received signal inter-symbol interference caused by the combination of TWTA non-linearity and band limiting filters. Although the gain measured is limited to about 0.4 db (Figure 11), the gain is largely dependent on the specific application scenario and will increase when using higher baud rates compared to the transponder bandwidth. Furthermore, in practical home based DTH implementations the likelihood of mismatches between the LNB and the satellite receiver is not negligible. Es/No Degradation in db 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 Acquisition without Equalisation Acquisition with Equalisation 0 QPSK 1/4 QPSK 1/3 QPSK 2/5 QPSK 1/2 QPSK 3/5 QPSK 2/3 QPSK 3/4 QPSK 4/5 QPSK 5/6 QPSK 8/9 QPSK 9/10 8PSK 3/5 8PSK 2/3 8PSK 3/4 8PSK 5/6 8PSK 8/9 8PSK 9/10 16APSK 2/3 16APSK 3/4 16APSK 4/5 16APSK 5/6 16APSK 8/9 16APSK 9/10 32APSK 3/4 32APSK 4/5 32APSK 5/6 32APSK 8/9 32APSK 9/10 Figure 11. Demodulator acquisition with and without equalisation. QEF thresholds are compared to E S /N 0 reference performance is the DVB-S2 standard. Equalisation is therefore very much recommended to improve the link performance, and could even be used to counter-act degradations coming from the terminal front end, allowing reducing the cost of the ODU and making the performance less installation dependent. Most demodulator implementations do include equalisers, with variable performances. F. Pilot Symbols Inserting the optional pilot symbols helps the demodulator channel estimation (acquisition and tracking) for more fragile modulation formats operating in presence of phase noise and/or modest signal to noise ratios. Alternatively pilots may be useful for very low C/(N+I) operations. In several configurations the use of pilots allows to lower the E S /N 0 threshold required for QEF behaviour. The overhead of pilot insertion is approximately 2.5% of the throughput (essentially, 36 pilot symbols every 16 SLOTS of 90 symbols = 1/40), which is not negligible for operations. The choice of pilot insertion should therefore be made on a case by case basis, depending on the physical link performance of each MODCOD with and without pilots. 9

Tests performed on the broadcast profile, with different phase noise conditions, have shown a significant advantage of pilot insertion (Figure 12) for QPSK-1/4, 8PSK-3/5 and 16APSK-2/3, where E S /N 0 acquisition was improved respectively by 0.3 db, 1.1 db and 0.8 db, representing an equivalent spectral gain increase of 8.1%, 10.7% and 7.7%. These MODCOD correspond to the lowest code rate of each modulation, where demodulator synchronisation conditions are the most challenging. No significant performance improvement has been measured for the other MODCOD. 1.2 Es/No Improvement in db 1 0.8 0.6 0.4 0.2 0 0.2 0.4 QPSK 1/4 QPSK 1/3 QPSK 2/5 QPSK 1/2 QPSK 3/5 QPSK 2/3 QPSK 3/4 QPSK 4/5 QPSK 5/6 QPSK 8/9 QPSK 9/10 8PSK 3/5 8PSK 2/3 8PSK 3/4 8PSK 5/6 8PSK 8/9 8PSK 9/10 16APSK 2/3 16APSK 3/4 16APSK 4/5 16APSK 5/6 16APSK 8/9 16APSK 9/10 32APSK 3/4 32APSK 4/5 32APSK 5/6 32APSK 8/9 32APSK 9/10 Figure 12. E S /N 0 acquisition improvement with pilots, compared to without pilots. Results are averaged from 3 different phase noise conditions: no phase noise, typical and critical masks of the DVB-S2 standard. Interestingly, the standard allows for the presence or absence of pilot symbols on a frame by frame basis. This can be useful in ACM mode where certain MODCOD may require a robust level of synchronisation. Taking measurements from Figure 12 as an example, the best settings would consist in activating pilot symbols only for the lowest code rates of QPSK, 8PSK and 16APSK modulations, and deactivating pilot symbols for the other MODCOD. It should be noted however that the capability to insert pilot symbols per MODCOD frame is currently not typically supported by current commercial DVB-S2 modulators. IV. Conclusion DVB-S2 performance has been characterised on three satellite platforms with many different configurations. The 28 MODCOD of the standard have successfully been passed by satellite. Performance degradations measured between L-band loopback in laboratory and real satellite conditions are acceptable when in the presence of realistic impairments. In the area of broadcasting, DVB-S2 allows for efficient transmission of HDTV channels with MPEG-4 video compression as well as for classical SDTV with MPEG-2 codec. The standard also allows for scalable broadcasting with the VCM mode, which can be coupled with SVC to avoid simulcasting. For interactive systems with a return link, ACM has shown very promising performances. It is a key feature for high throughput satellites systems (e.g. based on multi-beam Ka-band satellite), providing a significant increase of system capacity and availability compared to static systems. The important elements to consider for ACM systems are the accuracy of the E S /N 0 measurements by the demodulator, the ACM loop response time, the ACM margins, and handling of variable capacity links by the gateway. The interest of high order modulations has been demonstrated : when the link budget permits, and provided that TWTA pre-distortion and receiver equalisation are used, 16APSK and 32APSK modulations can push spectral efficiency up to 4.5 bit per symbol. In conclusion, DVB-S2 provides efficient solutions for a wide range of satellite applications such as television broadcasting, interactive broadband, professional applications and DSNG, with major improvements compared to the previous generations of the air interface. This contributes to the success of the standard, which has been adopted in many broadcast services worldwide, and in most of the recent broadband VSAT systems. 10

Acknowledgments The European Space Agency wishes to thank the industries, satellite operators and other participants who have been involved in the DVB-S2 Satellite Experiment project 4 which forms the basis of this publication. References 1 ETSI EN 302 307: "Digital Video Broadcasting (DVB); Second generation framing structure, channel coding and modulation systems for Broadcasting, Interactive Services, News Gathering and other broadband satellite applications", v1.2.1, Aug. 2009. 2 H. Bischl, H. Brandt, T. de Cola, R. De Gaudenzi, E. Eberlein, N. Girault, E. Alberty, S. Lipp, R. Rinaldo, B. Rislow, J. A. Skard, J. Tousch and G. Ulbricht, Adaptive coding and modulation for satellite broadband networks: From theory to practice, Int. Journal of Satellite Communications., John Wiley & Sons, Vol. 28, Mar./Apr. 2010, pp. 59-111. 3 N. Girault, E. Alberty, G. Verelst, C. Le Guern, R. De Gaudenzi, R. Rinaldo and O. Smeyers, Validation of DVB-S2 System Performances with ACM, Proceedings of the 27 th AIAA International Communications Satellite Systems Conference, Edinburgh, UK, Jun. 2009. 4 Final Report of ESA Contract No. 19572/06/NL/JA, DVB-S2 Satellite Experiment, EADS Astrium, Sept. 2010. 5 P. Hoarau, P. Inigo, N. Girault and C. Le Guern, Ka-band Measurements on DVB-S2 ACM System, Proceedings of the 15 th Ka and Broadband Communications, Navigation and Earth Observation Conference, Cagliari, Italy, Sept. 2009. 11