COST 280 -1- PM3-007 st 1 International Workshop L. Castanet - J. Lemorton - M. Bousquet July 2002 COST Action 280 “Propagation Impairment Mitigation for Millimetre Wave Radio Systems” A MODULAR SIMULATION TOOL OF INTERFERENCE AND FADE MITIGATION TECHNIQUES APPLIED TO MILLIMETER-WAVE SATCOM SYSTEMS ABSTRACT A simulation tool has been developed by ONERA to simulate the behaviour of the air interface of satellite communication systems operating at ku-band and above and more particularly to design and test Fade Mitigation Techniques (FMT) in order to improve link performances and Quality of Service (QoS) for the end-user. The objective of this contribution is to give a description of the functionality of this simulation tool and to illustrate it with some examples. 1 INTRODUCTION A simulation tool has been developed by ONERA in the framework of : the European action COST 255 “Radiowave propagation modelling for new SatCom services at Ku-band and above”, the RNRT project SAGAM of the French Ministry of Industry, and the European IST project GEOCAST. This tool enables the behaviour of the air interface of satellite communication systems operating at ku-band and above to be simulated and more particularly Fade Mitigation Techniques (FMT) to be designed and tested in order to improve link performances and Quality of Service (QoS) for the end-user. The objective of this contribution is to give a description of the functionality of this simulation tool and to illustrate it with some examples. In a first part a general description of the configurations implemented in to the simulator such as system architectures, FMT and interference issues. In a second part, the internal functions of the FMT control loop are described : firstly the detection schemes that can be simulated, secondly the short-time prediction of the propagation channel and thirdly a description of the decision process in terms of FMT activation thresholds and time delay issues. In a third part, typical statistical simulation results are presented, including example of control loop internal parameters optimisations and instantaneous performances and statistical results. L. Castanet, J. Lemorton Prof M. Bousquet ONERA SUPAERO DEMR / APR AECP 2 av. E. Belin - BP 4025 10 av. E. Belin - BP 4032 F-31055 Toulouse cedex 4 F-31055 Toulouse cedex Tel : +33.5.6225.2729 Tel : +33.5. 6217.8086 Fax : +33.5.6225.2577 Fax : +33.5. 6217.8345 E-m : email@example.com E-m : firstname.lastname@example.org, email@example.com COST 280 -2- PM3-007 st 1 International Workshop L. Castanet - J. Lemorton - M. Bousquet July 2002 2 SATCOM SYSTEM AIR INTERFACE SIMULATOR The simulator developed at ONERA [Castanet, December 2001] allows to simulate the performances of the air interface of SatCom systems operating in the 10 GHz – 50 GHz frequency range. In the current version of this simulator, the upper layers are considered only in terms of the constraints they impose on the FMT system configuration such as the detection scheme or the decision process. It means that the objective is not to simulate network performances or quality of service but more pragmatically quality of link. The simulation principle relies on a time-driven schedule, i.e. using a constant sampling period. The simulation is carried out with a sampling rate compatible with the variation of the propagation channel, that is at least several samples per second. This sampling rate allows to take into account the reaction time necessary for the system to adapt its configuration (transmitted power evolution, coding or modulation switches, signalling exchanges between components of the system, …) to the propagation channel fluctuations and not to evaluate fine performance like packet loss. 2.1 SatCom system architectures and performance assessment System architectures : Different system architectures can be taken into account through the implementation of their link budget. In particular, the modularity of the simulator allows regenerative and transparent configurations to be considered in the simulation. Up to now the performances of three systems have been assessed with the ONERA FMT simulator in different frameworks : • A Ka-band videoconferencing VSAT system characterised by a regenerative payload and a mesh VSAT network, studied in the framework of the European action COST 255 “Radiowave propagation modelling for new SatCom services at Ku-band and above [COST 255 Final Report, Chapter 6.2] [Mertens & Castanet, 2000], • A Ka-band ATM switch system aiming at providing multimedia application for the mass market using a regenerative payload and a meshed VSAT network, studied in the framework of the SAGAM project funded by the French Ministry of Industry [Castanet et al., October 2001], • A Ka-band packet switch system aiming at demonstrating the feasibility of multicast per satellite, using a regenerative payload and a star network, currently studied in the framework of the IST project GEOCAST [Castanet et al., September 2002]. As far as regenerative payloads are concerned, the implementation of the link budget into the simulator is relatively simple : it is performed only through the introduction of the link margins and of the Earth station receiver main characteristics for the downlink (in order to calculate the degradation of the figure-of merit of the Earth station in presence of clouds. If transparent configuration are of interest, the whole link budget has to be implemented in order to model the characteristics of the on-board TWTA. Current activity on the simulator aims first of all at introducing the whole link budget into the simulation in order to be able to take into account for interference perturbations, and at modelling the performances of a Ka-band bent-pipe system with a star network configuration. Performance and capacity criteria : The end-to-end performance of the link is evaluated from the introduction of propagation time series in the link budget. When the instantaneous BER of the received signal is higher than the COST 280 -3- PM3-007 st 1 International Workshop L. Castanet - J. Lemorton - M. Bousquet July 2002 minimum required objective (or when the instantaneous Eb/N0 is lower than the required one), it is considered that the link does not perform satisfactorily and consequently that an outage occurs. Figure 1 hereafter gives an example of the performance of the air interface of the GEOCAST system (right) [Castanet et al., September 2002] submitted to a strong propagation event (left) measured during the Olympus campaign by University Catholic of Louvain-la- Neuve (Belgium) [Vanhoenacker et al., 1990]. Figure 1 : Example of air interface performance without FMT On the basis of this error-rate criterion, a number of parameters are calculated to measure the efficiency of FMTs in the simulation. The main parameter categories are as follows : (i) Link Outage. The following parameters are computed : • number of outage higher than 10 s, lower than 10 s and total number of outage • duration of outage higher than 10 s, lower than 10 s and total duration of outage • fraction of time with outage higher than 10 s, lower than 10 s and total fraction of time with outage detected. (ii) Availability, that is the percentage of time for which the link is available. Based on Recommendation ITU-T G.826, the link is considered available if it operates properly (BER < 1.4 10-9) for more than ten consecutive seconds. The parameters considered in this study include : total number of availability period, total duration of availability period and ITU-T availability. (iii) Unavailability, that is the percentage of time for which the link is unavailable. In the same sense as availability parameter, the link is considered unavailable if an outage occurs (BER > 1.4 10-9 ) for more than ten consecutive seconds. The parameters under this category are : total number of unavailability period, total duration of unavailability period and ITU-T unavailability (iv) Duration of Return periods, that are defined as the time intervals between two unavailability periods. It differs from the availability period by the fact that it includes very short time intervals lasting for less than 10 seconds as well as very long time intervals comprising more one or more outages lasting for less than 10 seconds. The parameters are : total number of return period, total duration of return period and total fraction of time of return period COST 280 -4- PM3-007 st 1 International Workshop L. Castanet - J. Lemorton - M. Bousquet July 2002 (v) Mean duration, which includes the following parameters : mean duration of outage period, mean duration of availability period, mean duration of unavailability period and mean duration of return period Another set of parameters has been identified as relevant to characterise the behaviour of the system and to evaluate the efficiency of FMTs : (i) The percentage of time the system operates in each of the various level of power transmitted in case of ULPC (ii) The relative throughput denoted in percentage of the time, that is the ratio of the total number of bits effectively transmitted when the link is available to the maximum number of bits that would have been transmitted if the system operated in nominal conditions (without outage). This parameter allows an insight into the actual system capacity to be obtained. (iii) The percentage of time the system operates in each coding rate when adaptive coding is considered. As for the relative throughput, this parameter allows the actual system capacity to be estimated. (iv) The average switching rate, expressed in number per hour, defined as the ratio of the number of switches from one mode (SSPA output power, information data rate or coding rate) to another during the test sequence. 2.2 Fade Mitigation Techniques Among all FMT identified up to now [Willis and Evans, 1988] [Tartara, 1989] [Allnutt and Rogers, 1993] [Gallois, 1993] [Acosta, 1997] [Castanet et al., 1998], etc., it has been chosen to implement the most promising one according to the studied systems (see § 2.1). These FMT rely on the principles of Up-Link Power Control, Data Rate Reduction and Adaptive Coding. Up-Link Power Control (ULPC) : With ULPC, the output power of a transmitting Earth station is matched to up-link or down- link (in case of non-regenerative repeater) impairments. In the case of regenerative repeaters, up and down links budgets are independent, so ULPC acts only on the up-link budget. ULPC is used to keep a constant level of all the carriers at the input of the repeater, while maintaining the uplink budget close to target. Transmitter power is increased to counteract fade or decreased when more favourable propagation conditions are recovered so as to optimise satellite capacity. This FMT is simple to implement since it requires only the introduction of the minimum and the maximum power of the Earth station power amplifier, as well as the power increment. At this level it is important to recall that for this technique it is possible to play on the granularity (power increment) which is not possible so easily with other FMT. Data Rate Reduction (DRR) : Another technique implemented in the simulator consists in decreasing the information data rate at constant BER. The technique is called Data Rate Reduction. Here, user data rates should be matched to propagation conditions : nominal data rates are used under clear sky conditions (no degradation of the service quality with respect to the system margin), whereas reductions of data rates are introduced according to fade levels. COST 280 -5- PM3-007 st 1 International Workshop L. Castanet - J. Lemorton - M. Bousquet July 2002 As for ULPC, this technique is relatively easy to implement : it requires also to introduce the nominal data rate, the maximum acceptable degradation in terms of quality of service and the data rate variation step. Adaptive Coding (AC) : The introduction of redundant bits to the information bits when a link is experiencing fading, allows detection and correction of errors (FEC, …) caused by propagation impairments and leads to a reduction of the required energy per information bit. Adaptive coding consists in implementing a variable coding rate matched to impairments originating from propagation conditions. AC coding can be implemented in two ways : on the one hand alone which involves to be able to have the bandwidth vary accordingly to the extra coding and on the other hand in combination with DRR in order to work at constant bandwidth. Both solutions have been implemented in the ONERA FMT simulator. Joint FMT : Joint FMT [Castanet, December 2001] is a very promising solution to improve the performance of a SatCom system. The possibility to combine FMT has been introduced in the ONERA FMT simulator. Figure 2 (left) shows an example of FMT activation for the event given in § 2.1 : with ULPC, a combination of AC and DRR and a combination of the three techniques. The interest of such kind of combination clearly appears on these graphs. Figure 2 : examples of FMT activation, ULPC (left), ULPC+AC+DRR (right) 2.3 Interference issues With multibeam satellite communication systems, the level of interference impacts strongly of the performance of the physical layer [Sleight et al., 2001]. If interference is taken into account, it has a significant impact on the behaviour of the system submitted to strong propagation fading. Indeed, Figure 3 hereafter shows that the behaviour of the Eb/N0+I0 with and without interference differs especially for the area around the required Eb/N0. Of course, the link availability is lower with interference than without interference. In addition, the slope of the curves are also different, indicating a non-linear effect of the variation of the Eb/N0+I0 with respect to the time percentage and therefore with respect to the fading level (linear COST 280 -6- PM3-007 st 1 International Workshop L. Castanet - J. Lemorton - M. Bousquet July 2002 variation without interference for the uplink or without interference and without figure-of- merit degradation for the downlink). In the ONERA FMT simulator, three kinds of interference are being implemented : multi- beam interference and single-beam interference that are internal interference (that is due to components of the considered system), and adjacent system interference that are external interference (that is due to components belonging to another system). Figure 3 : example of up and down link budget from propagation loss prediction and interference calculation 3 INTERNAL FUNCTIONS OF THE FMT CONTROL LOOP The aim of a FMT control loop is to track the variations of the propagation channel in real time and to compensate propagation impairments either to increase its availability or to improve its instantaneous performance. For this purpose, it is first necessary to detect when a fade is occurring in order to assess if the quality of link is going to be degraded or if an outage is going to occur. Secondly, whenever an event supposed to lead to an outage is detected, it is necessary to check if the terminal is authorised to set up the mitigation, and upon reception of the clearance, to trigger the mitigation process. Another step can consist in performing a real time prediction of the propagation channel in order to compensate the reaction time of the system to obtain a better control loop behaviour. The following three functions are therefore implemented in the ONERA FMT simulator : the detection function, the prediction function and the decision function. 3.1 Detection schemes Two kinds of detection schemes are modelled in the simulator : on the one hand an open-loop ground detection scheme and on the other hand an hybrid-loop on-board detection scheme. The open-loop ground detection scheme relies on the estimation by the Earth station of the uplink (or downlink) impairment from a measurement of the propagation conditions. In the simulator, the open-loop ground detection scheme is based on the measurement of a downlink beacon (in general the satellite TT&R beacon) that could be in the downlink frequency band or in another band (for instance Ku-band). Once the downlink attenuated signal has been measured, the use of instantaneous frequency scaling algorithms between uplink and downlink frequencies allows the uplink fade to be estimated in real-time, for the downlink, COST 280 -7- PM3-007 st 1 International Workshop L. Castanet - J. Lemorton - M. Bousquet July 2002 the impairment is directly measured by the Earth station (description of this process in the next section). The interest of the hybrid-loop on-board detection scheme is on the one hand to avoid to implement a beacon receiver into the Earth station (especially for consumer or SOHO terminals for which the cost is a major constraint), and on the other hand to avoid to use frequency scaling algorithms (as in open-loop or in closed-loop detection schemes) thanks to a direct estimate of the uplink impairment. Here, the uplink impairment is estimated by the payload from a measurement of the power level of each individual carrier. Therefore, this detection scheme is implemented for regenerative repeaters only, for which it is possible to estimate the carrier power level directly on board. Indeed, as the signal is demodulated on- board the satellite, the Eb/N0+I0 ratio can be estimated in base-band and the transmitting Earth station can be identified from the signalling headers examination. 3.2 Short-time prediction of the propagation channel The detection concept defined previously can be considered is a "a posteriori" technique, since the system is able to react (through an appropriate FMT) only after detection of an event. Therefore, extra time delay is introduced when the detection of an event and the estimation of the necessary mitigation are carried out, in addition to the delay necessary for the activation of the FMT. These extra time delays lead to error contributions, that could be reduced through the implementation of short-term predictions into the FMT control loop. The architecture of the propagation channel short-time predictor is given at Figure 4 hereafter. Figure 4 : Architecture of the FMT control loop predictor 3.2.1 Attenuation prediction The attenuation prediction follows two or three consecutive steps : first of all a filtering of the fast fluctuating component of the monitored signal, afterwards a frequency scaling of the monitored signal (in the open loop scheme only) and finally a real time prediction. The filtering is performed with different types of filters : moving average window, rectangular filter, exponential filter, raised cosine filter, butterworth… The input parameters are the number of coefficients of the filter as well as the cut-off frequency. As far as frequency scaling is concerned, several methods have been implemented in the simulator such as : exponential, ITU-R, COST 205 (fixed and variable), Hodge, Laster- Stutzman, Rücker, or Gremont EFSR and IFSR methods (see the COST 255 final report). Regarding the short-term prediction method, only a rough technique relying on the estimation of the instantaneous fade slope has been implemented up to now. However, the modular COST 280 -8- PM3-007 st 1 International Workshop L. Castanet - J. Lemorton - M. Bousquet July 2002 architecture of the simulator allows more sophisticated prediction techniques to be implemented if necessary. 3.2.2 Scintillation variable margin As scintillation is unpredictable in real time, a predictor of the envelope of the scintillation component relying on a variable detection margin has been proposed by UCL [Mertens & Castanet, 2000] and implmented in the simulator (see Figure 4). The objective of this prediction method is to add to the predicted attenuation a variable margin, which corresponds to the slow varying envelope of the scintillation (due to temperature, humidity and cloud integrated liquid water content). After having filtered out the fast varying component of the monitored signal, its log-amplitude is loaded into a FIFO shift register containing the last 21 samples upon which standard deviation is returned. To estimate the severity of scintillation log-amplitudes at the transmission link frequency, the expected joint stochastic process of clear-air amplitude scintillation along with rapid fluctuations of rain attenuation, supposed to be approximately gaussian with zero-mean and with a RMS intensity, is estimated by moving standard deviation of past fluctuation amplitudes of the monitored signal. If needed (open-loop scheme) a frequency scaling law can be applied on the estimated scintillation standard deviation (see next paragraph). Scintillation fades at the higher frequency are therefore statistically compensated by biasing the rain fade predictor with the scintillation amplitude level not exceeded for some prescribed percentage of time (scintillation signal envelope definition). 3.3 Decision process When a fade is occurring, a primary concern is to initiate in due time the appropriate compensation. The control algorithm implemented in this simulator makes use of a detection margin. The aim of such a detection margin in a FMT control loop allows to prevent the system from : firstly prediction errors arising in the detection / prediction process, secondly errors due to the frequency scaling method (open-loop scheme) and thirdly disturbances due to the fast fluctuating component of the received signal. Figure 5 : example of optimisation of a FMT control loop from [Castanet, 2001] When more favourable propagation conditions are recovered, FMT are gradually disabled and the nominal operating mode is restored. An additional hysteresis margin HM is employed to prevent repeated mode switching if the predicted signal fluctuates around the fade detection COST 280 -9- PM3-007 st 1 International Workshop L. Castanet - J. Lemorton - M. Bousquet July 2002 threshold. Another way to take into account the hysteresis effect would consist in introducing a hysteresis delay, which is not yet implemented in the simulator. Figure 5 presents an example of sensitivity analysis carried out with respect to detection and hysteresis margins. Time delay issues is also an important problem to deal with. Different time delays are taken into account in the simulator : the first contribution to the delay budget is due to the detection phase (single or double hop depending on the detection scheme), the second one is the delay corresponding to the sampling rate of the measurement, the third one is the filtering delay (that depends on the number of coefficients). As far as FMTs are concerned, depending if the authorisation of the Network Control Centre has to be requested, a new delay is introduced as well as for the reaction time of the DAMA, which has to wait for the upgrade of its connection control matrix. 4 TYPICAL SIMULATION RESULTS Examples of simulation results obtained with this simulator have been presented in previous sections (see Figures 1 to 5) to perform FMT control loop optimisation and event-based analyses. Figure 6 hereafter shows an example of statistical analysis performed from the introduction of several months of propagation data. ITU-T availability System Throughput (%) (%) 100 100 99.5 99 A u v t 98 99 a i N 97 i 98.5 l o l i 96 a 98 z No FMT F 95 b a ULPC ULPC M i 97.5 t 94 UL DRR UL DRR T l i 97 93 both DRR t both DRR o y 96.5 n 92 96 91 90 95.5 1 1 3 3 5 5 7 7 10 10 Month 11 Month 11 Figure 6 : example of long-term analysis performed in the framework of COST 255 5 CONCLUSION AND FURTHER WORK A simulation tool has been developed by ONERA in the framework of the projects : COST 255, SAGAM and GEOCAST. This tool enable the behaviour of the air interface of satellite communication systems operating at ku-band and above to be simulated and more particularly Fade Mitigation Techniques (FMT) to be designed and tested in order to improve link performances and Quality of Service (QoS) for the end-user. This simulation tool allows different link budgets to be taken into account and both statistical and dynamics analyses to be performed in order to design FMT and optimise system performances. Statistical analysis enables the interest to implement FMT to be assessed through estimates of propagation impairments that the system will have to face within the coverage area of interest or system availability with respect to different FMT schemes. Dynamics analysis enables the performances of a SatCom system to be studied through the introduction of propagation time series. These performances are estimated in terms of link availability (outage, ITU-T availability, ITU-T unavailability, …) and in terms of parameters COST 280 - 10 - PM3-007 st 1 International Workshop L. Castanet - J. Lemorton - M. Bousquet July 2002 having a strong impact on the system behaviour (for instance number of switches between states) or on the system capacity (throughput as information data rate supplied to the end- user). This simulation tool has been developed on a modular way, it is therefore possible to upgrade it with new models. At the moment, different configurations can be simulated such as : Ku, Ka or EHF systems, transparent or regenerative payloads, single or multispot beams. Different FMT can be simulated such as Up-Link Power Control, Data Rate Reduction, Adaptive Coding or Adaptive Modulation. For a given FMT, different control loop configuration can be taken into account such as open-loop, closed-loop or hybrid-loop with ground or on-board detection. The key-actions for the on-going activity on this simulator in the framework of COST 280 will be to address interference and signalling issues. Once these aspects will have been dealt with, other FMT such as Site Diversity, Frequency Diversity or Time Diversity could be implemented into the simulator. In addition, new short-term prediction models are due to be developed in order to improve the accuracy of the control loop. VI REFERENCES Acosta R.J. : "Rain fade compensation alternatives for Ka-band communication satellites", 3rd Ka-band Utilization Conference, Sorrento, Italy, 15-18 Sept. 1997. Allnutt J.E. - Rogers D.V. : "Recent developments in propagation counter-measures for VSAT services", ICAP’93, Heriot-Watt University, UK, 30 March - 2 April 1993. Castanet L., Lemorton J., Bousquet M. : "Fade Mitigation techniques for New SatCom services at Ku-band and above : a Review", Fourth Ka-band Utilization Conference, Venice, 2-4 November 1998. Castanet L., Lemorton J., Bousquet M., Garnier B. : "Illustration of the interest to use propagation information to design Fade Mitigation Techniques : application to the SAGAM system", COST 280 2nd Management Commitee Meeting, doc. PM2-011, Toulouse, France, 29-30 October 2001. Castanet L. : "Fade Mitigation Techniques for new SatCom systems operating at Ka and V bands", Ph’D of SUPAERO, Toulouse, France, 18 Decembre 2001. Castanet L., Lemorton J., Bousquet M., Claverotte L. : "A joint Fade Mitigation Technique applied to the regenerative packet switch payload of the GEOCAST system", 8th Ka-band Utilization Conference, Baveno, Italy, 25-27 September 2002. COST 255 : "Ka-band videoconference VSAT system", COST 255 Final Report, Chapter 6.2, to be published. Gallois A.P. : "Fade countermeasure techniques for satellite communication links", Int. Symp. on Comms Theory and Applications, July 1993. Mertens D., Castanet L. : "Performance simulation of an adaptive data rate scheme for rain fade compensation in a Ka-band VSAT videoconferencing system", HF (Belgian Journal of Electronics & Communications), nº1, 2000. Sleight S., Toth F., Thorburn M., Fashano M. : "Parametric report of interference in Ka-band multiple-beam broadband satellite payload architectures", 7th Ka-band Utilisation Conference, Portofino, Italy, 26-28 September 2001. Tartara G. : "Fade countermeasures in millimetre-wave satellite communications : a survey of methods and problems", Proc. Olympus Util. Conference, Vienna, Austria, April 1989. Vanhoenacker D., Matagne J., Vyncke C., Vander Vorst A. : "Preliminary Results of the Belgian Olympus Experiment," presented at the 13th Meeting of Olympus propagation Experimenters, ESTEC, Noordwijk, March 1990, pp. 105-122. Willis M.J. - Evans B.G. : "Fade countermeasures at Ka-band for OLYMPUS", Int. Jour. of Sat. Com., Vol. 6, June 88, pp. 301-311.